A GIS-based urban land planning analysis method and system

Abstract

The present application relates to the technical field of geographic information system, and more particularly to a kind of city land planning analysis method and system based on GIS, by integrating remote sensing, historical planning, mobile terminal and social media data, construct multi-scale spatiotemporal cube model, generate standardized city temporary land data, adopt adaptive feature extraction to identify temporary land, establish hierarchical classification system, and monitor abnormal change through time series model, construct the association diagram of temporary land and city function network, analyze interactive relationship, finally propose differentiated city planning optimization suggestion, help city land resource efficient allocation and management, by fusing multi-source data, realize the multidimensional cognition of space, time, function and relationship, fundamentally improve the understanding depth of temporary land.

Description

Technical Field

[0001] This invention relates to the field of urban land planning technology, and in particular to a GIS-based urban land planning analysis method and system. Background Technology

[0002] GIS (Geographic Information System) data is the foundation of GIS technology. It includes geospatial and attribute data, used for storing, analyzing, and displaying geographic information. GIS technology enables the collection and integration of geographic information such as topography and land types, providing basic data for land use planning. It can analyze the spatial relationships between land parcels with different uses and simulate and analyze land planning schemes through model building. GIS technology can quickly and accurately analyze and comprehensively evaluate the spatial distribution and characteristics of land, providing a scientific basis and decision support for land planning.

[0003] Land use patterns are considered a crucial factor influencing the balance between work and residence for residents, which in turn affects urban residents' transportation, and even their carbon emissions and the urban environment. Among related technologies, analyzing urban land use vector data from different periods using software such as ArcGIS can determine the overall characteristics of the land, regional differentiation patterns, and changes in major land use types.

[0004] However, traditional urban planning often differs from the actual urban spatial situation. Traditional land survey data typically focuses on permanent building land, lacking analysis of temporary building land, making it difficult to accurately reflect the true urban spatial situation and apply it to urban land planning and analysis. A Chinese invention patent, CN118761560B, entitled "A GIS-based Urban Land Planning Analysis Method and System," includes: identifying temporary building land blocks and their corresponding temporary building land types in a target city; determining the geospatial variability of temporary building land blocks based on first and second historical geospatial data; determining whether temporary building land blocks are anomalous based on historical and current remote sensing images of the target city at the second historical time point; and performing correlation analysis between temporary building land blocks and land blocks of the same type. By analyzing current remote sensing images of the target city, temporary building land blocks and their corresponding temporary building land types can be identified, and the attribute information of temporary building land blocks marked as anomalous can be used as a reference for analysis of the same land type. In the process of rapid urbanization, temporary building land, as an important part of urban space, often faces the following challenges in its planning and management:

[0005] 1. Single data source: Existing technologies rely too heavily on a single type of remote sensing image data, which is easily affected by weather conditions and image quality, resulting in insufficient accuracy in identifying temporary buildings.

[0006] 2. Insufficient timeliness: There is a lack of a continuous monitoring mechanism for changes in land use for temporary buildings, making it difficult to detect cases of illegal or overdue use in a timely manner.

[0007] 3. Limited analytical dimensions: Traditional methods analyze temporary buildings as isolated entities, ignoring their complex interactions with the urban functional network.

[0008] 4. Insufficient ability to predict changes: Existing technologies mainly focus on changes that have already occurred and lack the ability to predict the future evolution of land used for temporary buildings.

[0009] 5. Insufficient decision support: There is a lack of effective mechanisms to translate analysis results into actual planning actions, and planning recommendations often remain merely formalities.

[0010] Therefore, there is an urgent need for an urban land planning analysis method and system that can comprehensively utilize multi-source data, deeply analyze the land change patterns of temporary buildings, reveal their relationship with urban functions, and provide effective decision support. Summary of the Invention

[0011] In view of the shortcomings of the prior art, the purpose of this invention is to provide a GIS-based urban land planning analysis method and system. Through multi-source heterogeneous data fusion, adaptive feature extraction, temporal change analysis, functional correlation mining, and planning optimization suggestion generation, it can achieve comprehensive analysis and intelligent planning of urban temporary land use, thereby improving the management efficiency and planning scientificity of urban land resources.

[0012] The technical solution of this invention is: a GIS-based urban land planning analysis method, comprising: acquiring multi-source heterogeneous spatiotemporal data, including remote sensing image data, historical planning data, mobile terminal data, and social media geotagged data; constructing a unified multi-scale spatiotemporal cube model based on the acquired multi-source heterogeneous spatiotemporal data, and generating standardized multimodal urban temporary land use characterization data; identifying urban temporary land use areas using an adaptive feature extraction mechanism based on the standardized multimodal urban temporary land use characterization data, and constructing a hierarchical classification system for temporary land use; establishing a change-sensitive time series model based on the hierarchical classification system for temporary land use and historical time series data, detecting and identifying temporary land use areas with abnormal change patterns; constructing a correlation graph model between temporary land use and urban functional networks based on the identified temporary land use areas with abnormal change patterns, and analyzing the interaction relationship between abnormal temporary land use and urban functional systems; and generating differentiated urban planning optimization suggestions based on the interaction relationship and multi-objective evaluation indicators to support the efficient allocation and management of urban land resources.

[0013] Preferably, the acquisition of multi-source heterogeneous spatiotemporal data includes:

[0014] Acquire optical and radar remote sensing image data of the target urban area from a remote sensing satellite platform;

[0015] Retrieve historical planning data and land use information for the target urban area from the urban planning database;

[0016] Collect ground-based real-view data and location information of the target urban area using mobile terminal devices;

[0017] Extract geotagged data and user activity information related to target city areas from social media platforms;

[0018] The quality of the acquired data is assessed, and a data reliability scoring system is constructed based on information entropy and texture features to determine the weight coefficients of each data source.

[0019] Preferably, the construction of a unified multi-scale spatiotemporal cube model includes:

[0020] Spatiotemporal alignment of multi-source heterogeneous data is performed based on feature point matching algorithm to ensure consistency of data from different sources in time and space dimensions;

[0021] Construct a four-dimensional spatiotemporal data structure that integrates spatial coordinates (x, y, z) and the time dimension (t) into a unified data representation model;

[0022] Scale transformation is performed on spatial data of different resolutions to achieve unified management of multi-scale spatial information;

[0023] Interpolation is performed on discontinuous time series data to establish continuous and equally spaced time series data.

[0024] Perform semantic-level fusion and standardization of multi-source data to ensure semantic consistency of data from different sources.

[0025] Preferably, the method of identifying urban temporary land use areas using an adaptive feature extraction mechanism includes:

[0026] A multi-scale feature extraction network is constructed to extract hierarchical features from urban area images;

[0027] By integrating spectral features, texture features, morphological features, and contextual features, a multidimensional feature representation of temporary land use is constructed.

[0028] Optimize feature weight allocation based on attention mechanism to highlight key distinguishing features of temporary land use;

[0029] Accurate boundary extraction of temporary land use areas is achieved through a context-aware segmentation algorithm.

[0030] The extracted temporary land use areas are labeled with attributes, including information such as area, shape, material, and purpose.

[0031] Preferably, the hierarchical classification system for temporary land use includes:

[0032] Establish a three-level classification framework, including a category layer (temporary building types), a function layer (usage function), and a purpose layer (specific purpose);

[0033] For the category level, temporary land use areas are divided into temporary building land, temporary facility land, temporary activity land, and temporary storage land;

[0034] For the functional layer, temporary use areas are divided into production functions, living functions, service functions, and public functions;

[0035] For the application layer, it is further subdivided into specific applications such as emergency rescue and disaster relief, engineering construction, commercial exhibitions and sales, and cultural activities, based on specific application scenarios.

[0036] Based on weakly supervised learning, a classification model is trained using a small number of labeled samples to achieve automatic classification of large-scale temporary land use.

[0037] Preferably, the establishment of the change-sensitive time series model includes:

[0038] Based on the planning cycle and expected duration of temporary land use, construct time window parameters;

[0039] Design a loss function that is sensitive to changes in the characteristics of temporary land use to improve the detection sensitivity of changes in small targets and sparsely distributed temporary buildings;

[0040] A bidirectional time-series encoder was constructed to encode and analyze the epochal change sequence of temporary land use;

[0041] By introducing an urban planning rule base, anomaly detection is constrained by domain knowledge, thereby reducing the false judgment rate;

[0042] A multi-evidence fusion method based on evidence theory is used to quantify the reliability of anomaly detection and generate a confidence score.

[0043] Preferably, the temporary land use area for detecting and identifying abnormal change patterns includes:

[0044] The temporal variation curves of temporary land use areas are fitted to detect outliers that deviate from the expected change patterns;

[0045] Anomaly scores are calculated based on the rate of change and the magnitude of change, and temporary land use areas that exceed the preset threshold are marked as potential anomalies.

[0046] By combining urban planning cycles and land use conversion rules, the authenticity of potential abnormal areas can be verified;

[0047] The identified abnormal temporary land use areas are classified into different types, such as those not demolished on schedule, those undergoing functional transformation, and those requiring structural reinforcement.

[0048] Generate a spatial distribution heat map of abnormal temporary land use to visually demonstrate the clustering of abnormal areas.

[0049] Preferably, the construction of the correlation graph model between temporary land use and urban functional network includes:

[0050] Based on the spatial relationship between temporary land use and surrounding permanent buildings, construct a heterogeneous spatial relationship diagram;

[0051] Calculate the correlation strength between temporary land use and surrounding urban functional nodes, including indicators such as spatial proximity, functional complementarity, and frequency of pedestrian interaction;

[0052] A multi-layer graph convolutional network is used to extract the structural and functional features of temporary land use nodes;

[0053] Temporary land use communities with similar functional patterns and evolution trends were discovered based on spectral clustering algorithms;

[0054] Analyze the location and importance of abnormal temporary land use within the urban functional network, and assess its value for preservation or redevelopment.

[0055] Preferably, the generation of differentiated urban planning optimization suggestions includes:

[0056] Construct a multi-objective evaluation index system that includes economic benefits, social value, and environmental impact;

[0057] Based on cellular automata and system dynamics models, we simulate urban evolution scenarios under different planning strategies.

[0058] Based on the characteristics of abnormal temporary land use and the needs of urban functions, generate personalized planning suggestions, including demolition and reconstruction, functional upgrades, and structural modifications.

[0059] Provides visualization and analysis tools to intuitively display the expected effects and impact assessments of planning recommendations;

[0060] Design a collaborative decision-making mechanism involving multiple parties, integrate feedback from planners, citizens, and developers, and optimize planning schemes.

[0061] The present invention also provides a GIS-based urban land planning and analysis system, including a processor and a memory; wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the GIS-based urban land planning and analysis method as described above.

[0062] This invention innovatively combines multi-source heterogeneous data fusion, spatiotemporal modeling, deep feature extraction, temporal anomaly detection, and correlation network analysis to construct a complete closed-loop system for temporary land use analysis and planning. Compared with existing technologies, it has the following advantages and beneficial effects:

[0063] 1. Data multidimensionality: By integrating multi-source data, a multidimensional understanding of space, time, function, and relationship is achieved, fundamentally improving the depth of understanding of temporary land use.

[0064] 2. Identification accuracy: The adaptive feature extraction mechanism can accurately identify and classify various types of temporary land use, with an identification accuracy that is more than 30% higher than that of traditional methods.

[0065] 3. Anomaly Sensitivity: Change-sensitive time series models are highly sensitive to abnormal change patterns in temporary land use and can promptly identify temporary buildings that have not been demolished after the planning period.

[0066] 4. Relationship Insight: By using graph neural networks, we reveal the complex interaction between temporary land use and urban functional networks, elevating point-based analysis to network-based understanding.

[0067] 5. Forward-looking decision-making: Based on multi-objective evaluation and scenario simulation, it provides differentiated planning optimization suggestions, realizing a paradigm shift from passive response to proactive planning.

[0068] Overall, this invention, through technological innovation, has transformed the analysis of temporary land use from a static, isolated, and passive traditional paradigm to a dynamic, networked, and proactive modern paradigm, providing a new perspective and powerful tools for urban land planning. Attached Figure Description

[0069] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below.

[0070] Figure 1 This is a flowchart illustrating a GIS-based urban land planning analysis method in one embodiment of the present invention.

[0071] Figure 2 This is a schematic diagram of multi-source heterogeneous spatiotemporal data acquisition in one embodiment of the present invention.

[0072] Figure 3 This is a schematic diagram of the structure of a multi-scale spatiotemporal cube model in one embodiment of the present invention.

[0073] Figure 4 This is a flowchart of the adaptive feature extraction mechanism in one embodiment of the present invention.

[0074] Figure 5 This is a schematic diagram of the hierarchical classification system for temporary land use in one embodiment of the present invention.

[0075] Figure 6 This is a flowchart of the abnormal temporary land use detection process in one embodiment of the present invention.

[0076] Figure 7 This is a structural block diagram of a GIS-based urban land planning and analysis system according to one embodiment of the present invention. Detailed Implementation

[0077] 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. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0078] Example 1: Overall Process of GIS-Based Urban Land Planning Analysis Method

[0079] like Figure 1 As shown in the figure, the GIS-based urban land planning analysis method provided by this invention mainly includes the following steps:

[0080] Step S1: Acquire multi-source heterogeneous spatiotemporal data, including remote sensing image data, historical planning data, mobile terminal data, and social media geotagging data.

[0081] Step S2: Based on the acquired multi-source heterogeneous spatiotemporal data, construct a unified multi-scale spatiotemporal cube model and generate standardized multimodal urban temporary land use characterization data.

[0082] Step S3: Based on the standardized multimodal urban temporary land use characterization data, an adaptive feature extraction mechanism is used to identify urban temporary land use areas and construct a hierarchical classification system for temporary land use.

[0083] Step S4: Based on the hierarchical classification system and historical time series data of the temporary land use, establish a change-sensitive time series model to detect and identify temporary land use areas with abnormal change patterns.

[0084] Step S5: Based on the identified temporary land use areas with abnormal change patterns, construct a graph model of the relationship between temporary land use and urban functional networks, and analyze the interaction between abnormal temporary land use and urban functional systems.

[0085] Step S6: Based on the interaction relationship and multi-objective evaluation indicators, generate differentiated urban planning optimization suggestions to support the efficient allocation and management of urban land resources.

[0086] Through the above steps, the method of this invention achieves a complete closed loop from data acquisition, spatiotemporal modeling, feature extraction, anomaly detection, correlation analysis to decision support, establishing a scientific and comprehensive framework for temporary land use analysis and planning. This method organically integrates multi-source heterogeneous data, enabling not only accurate identification of temporary land use but also the discovery of abnormal change patterns. Furthermore, it provides targeted planning suggestions in conjunction with urban functional networks, thereby improving the efficiency of urban land resource management and the scientific nature of planning.

[0087] Example 2: Acquisition of Multi-Source Heterogeneous Spatiotemporal Data

[0088] In this embodiment, as Figure 2 As shown, the specific steps for obtaining multi-source heterogeneous spatiotemporal data include:

[0089] First, optical and radar remote sensing imagery data of the target urban area are acquired from remote sensing satellite platforms. Optical remote sensing data primarily includes high-resolution satellite imagery, such as sub-meter resolution imagery provided by QuickBird and WorldView series, as well as medium-resolution Landsat and Sentinel series imagery. Radar remote sensing data includes SAR (Synthetic Aperture Radar) imagery, which has all-weather, day-and-night observation capabilities, compensating for the limitations of optical imagery in cloudy or foggy weather. Preferably, imagery data with a spatial resolution in the range of 0.5–5 meters are selected to ensure effective identification of temporary structures.

[0090] Secondly, historical planning data and land use information for the target city area are retrieved from the urban planning database. Historical planning data includes land use planning maps from different periods, urban master plans, regulatory detailed plans, and various special plans. Land use information includes attribute data such as land use nature, planning period, and construction intensity. This data is typically stored in the spatial database of the urban planning management department and has high reliability and authority.

[0091] Third, ground-based real-scene data and location information for the target urban area are collected using mobile terminal devices. Ground-based real-scene data is primarily acquired through dedicated data collection applications on mobile terminals, including ground photos, videos, and 3D point cloud data, along with precise geographic location information and capture time. Preferably, a distributed mobile data collection network is constructed to integrate data collected by urban management personnel, professional surveyors, and volunteers, forming a comprehensive ground-based real-scene database.

[0092] Fourth, extract geotagged data and user activity information related to the target urban area from social media platforms. This includes social media text, images, and video content with geotags, reflecting public usage and subjective feelings about urban spaces. Preferably, semantic analysis and computer vision technologies are used to extract information related to temporary land use from social media data, such as building appearance, usage status, and public evaluation.

[0093] Finally, the quality of the acquired data was assessed, and a data reliability scoring system was constructed based on information entropy and texture features to determine the weight coefficients of each data source. The data reliability scoring system comprehensively considers factors such as data timeliness, spatial accuracy, completeness, and consistency, providing a scientific basis for subsequent data fusion.

[0094] The method for calculating information entropy to assess data reliability is as follows:

[0095]

[0096] Where H(X) represents the information entropy of data source X, p(x) i ) represents data element x i The probability is given by , where n represents the total number of data elements. Higher information entropy indicates greater uncertainty in the data, requiring more processing and verification.

[0097] Based on information entropy and other evaluation metrics, the weighting coefficients for each data source are calculated:

[0098]

[0099] Among them, W i H represents the weight coefficient of the i-th data source. i Let Q represent the normalized information entropy of the i-th data source. i The quality score of the data source is represented by a value ranging from 0 to 1, which represents a comprehensive score of data integrity, timeliness, and reliability. α is an adjustment parameter (usually ranging from 0 to 1, used to control the influence of information entropy in weight calculation), and m is the total number of data sources.

[0100] Through the above steps, comprehensive acquisition and quality assessment of multi-source heterogeneous spatiotemporal data were achieved, laying a solid foundation for subsequent data fusion and analysis. Different types of data complement each other, collectively forming a multi-dimensional observation network for urban temporary land use, significantly improving the comprehensiveness and reliability of the data.

[0101] Example 3: Construction of a Multi-Scale Spatiotemporal Cube Model

[0102] In this embodiment, as Figure 3 As shown, the specific steps for constructing a unified multi-scale spatiotemporal cube model include:

[0103] First, spatiotemporal alignment of multi-source heterogeneous data is performed based on a feature point matching algorithm to ensure consistency in time and space dimensions. Feature point matching employs an improved SIFT (Scale Invariant Feature Transform) algorithm combined with a Spatial Transform Network (STN) to achieve accurate registration of heterogeneous data. Preferably, for the registration of remote sensing imagery and ground scene photographs, a multi-view geometrically constrained feature matching method is used, achieving sub-pixel accuracy.

[0104] The key step in feature point matching is calculating the similarity between feature descriptors, using cosine similarity:

[0105]

[0106] Where S(A,B) represents the similarity between feature descriptors A and B, A i and B i Let A and B represent the i-th components of feature descriptors A and B, respectively, and n represent the dimension of the feature descriptors. The closer the similarity value is to 1, the higher the matching degree between the two feature points.

[0107] Secondly, a four-dimensional spatiotemporal data structure is constructed, integrating spatial coordinates (x, y, z) with the time dimension (t) into a unified data representation model. This four-dimensional spatiotemporal data structure employs an organization method combining an octree and a time axis, supporting multi-resolution spatial data storage and efficient retrieval. Preferably, the storage granularity of the spatiotemporal data can be adaptively adjusted according to the importance and frequency of change of urban areas, with finer spatial resolution and time intervals used for key areas and frequently changing areas.

[0108] Third, scaling is performed on spatial data of different resolutions to achieve unified management of multi-scale spatial information. The scaling method combines wavelet transform and pyramid structure to maintain the integrity of information at different scales. Preferably, a 5-layer scale pyramid is constructed, with the lowest layer having a resolution of 0.5 meters and the resolution ratio between adjacent layers being 2:1, to achieve multi-scale representation from architectural details to the overall urban landscape.

[0109] Fourth, interpolation is performed on discontinuous time series data to establish continuous and equally spaced time series data. Time series interpolation employs a multidimensional interpolation method based on tensor decomposition, which can simultaneously consider the correlation between spatial and temporal dimensions. Preferably, linear interpolation, spline interpolation, or deep learning-based interpolation methods are adaptively selected based on the temporal distribution characteristics of the data to ensure the accuracy of the interpolation results.

[0110] Finally, semantic-level fusion and standardization are performed on multi-source data to ensure semantic consistency across different sources. Semantic fusion employs ontology mapping and knowledge graph technologies to establish semantic relationships between different data sources. Preferably, a temporary land use domain ontology model is constructed, comprising three levels: category, attribute, and relationship, to achieve semantic unification across data sources.

[0111] Through the steps described above, this embodiment constructs a unified multi-scale spatiotemporal cube model, transforming multi-source heterogeneous data into standardized multimodal urban temporary land use representation data. This model not only achieves spatiotemporal and scale uniformity of the data but also establishes semantic-level associations and mappings, laying the foundation for subsequent temporary land use feature extraction and analysis.

[0112] Example 4: Adaptive Feature Extraction Mechanism

[0113] In this embodiment, as Figure 4 As shown, the specific steps for identifying temporary urban land use areas using an adaptive feature extraction mechanism include:

[0114] First, a multi-scale feature extraction network is constructed to extract hierarchical features from urban area images. The multi-scale feature extraction network adopts a Feature Pyramid Network (FPN) architecture, combined with a Deep Residual Network (ResNet) as the backbone network, to achieve effective extraction from low-level features to high-level semantic features. Preferably, the network depth is set to 50 layers, containing 5 feature scales, which can simultaneously capture the detailed features and contextual information of temporary buildings.

[0115] The core of the multi-scale feature pyramid network is the feature fusion mechanism. The top-down and bottom-up feature transfer can be represented as:

[0116] P l =Conv(UpSample(P) l-1 )+Lateral l ),

[0117] Among them, P l Represents the feature map of layer l, UpSample indicates the upsampling operation (usually nearest neighbor interpolation or bilinear interpolation, used to increase the feature map resolution by a factor of 2), Lateral l represents the lateral connection features of layer l (obtained by 1×1 convolution of features from the corresponding layer of the backbone network), and Conv represents the convolution operation (usually a 3×3 convolution, used to eliminate the aliasing effect caused by upsampling). This structure can effectively combine high-level semantic information with low-level positional information.

[0118] Secondly, a multi-dimensional feature representation of temporary land use is constructed by integrating spectral features, texture features, morphological features, and contextual features. Spectral features mainly utilize the reflectance characteristics of different land features in various bands; texture features are extracted using gray-level co-occurrence matrix and local binary mode; morphological features include geometric descriptors such as area, perimeter, and compactness; and contextual features consider the relationship between the target and its surrounding environment. Preferably, for the identification of temporary buildings, the focus is on material texture, regularity, and temporal variation characteristics, which are key indicators for distinguishing temporary buildings from permanent buildings.

[0119] Third, feature weight allocation is optimized based on an attention mechanism to highlight key distinguishing features of temporary land use. The attention mechanism combines channel attention and spatial attention to adaptively adjust feature importance. Channel attention focuses on specific features, while spatial attention focuses on specific locations; this combination enables precise localization of key feature regions. Preferably, the attention module employs a lightweight design, keeping computational complexity within 5% of the original network, thus improving performance while maintaining computational efficiency.

[0120] The calculation process of the channel attention mechanism is as follows:

[0121] M c (F)=σ(MLP(AvgPool(F))+MLP(MaxPool(F))),

[0122] Among them, M c The channel attention map (with a value range of [0,1], representing the importance weight of each channel) is defined by the following: F represents the input feature map; AvgPool and MaxPool represent global average pooling and max pooling, respectively (compressing the feature map into a 1×1×C vector, where C is the number of channels); MLP represents a multilayer perceptron (typically containing two fully connected layers with a dimensionality reduction layer in between); and σ represents the sigmoid activation function. The channel attention map has a value range of [0,1], representing the importance weight of each channel.

[0123] Fourth, a context-aware segmentation algorithm is used to accurately extract the boundaries of temporary land use areas. This algorithm, based on the DeepLabv3+ architecture, combines dilated convolutions and an encoder-decoder structure to preserve spatial detail while maintaining the receptive field size. Preferably, a Conditional Random Field (CRF) is used as the post-processing module to further optimize the smoothness and accuracy of the segmentation boundaries.

[0124] Finally, the extracted temporary land use areas are labeled with attributes, including area, shape, material, and purpose. Attribute labeling employs a multi-task learning framework, simultaneously performing semantic segmentation and attribute classification to improve feature utilization efficiency. Preferably, dedicated branch networks are set up for different attributes, such as a material recognition branch and a purpose classification branch. Each branch shares the features of the backbone network but has an independent output layer.

[0125] Through the steps described above, this embodiment achieves accurate identification and attribute labeling of urban temporary land use areas. The adaptive feature extraction mechanism can flexibly adjust the feature extraction strategy according to data characteristics and task requirements, significantly improving the accuracy and robustness of temporary land use identification and laying the foundation for subsequent classification and anomaly detection.

[0126] Example 5: Construction of a Hierarchical Classification System for Temporary Land Use

[0127] In this embodiment, as Figure 5 As shown, the specific steps for constructing a hierarchical classification system for temporary land use include:

[0128] First, a three-level classification framework is established, including a category layer (temporary building types), a function layer (usage function), and a purpose layer (specific use). This three-level framework employs a hierarchical organizational structure, with each level refining and specifying the previous one. Preferably, the classification system uses a top-down design approach, first establishing a conceptual model, then instantiating and validating it to ensure the scientific validity and practicality of the classification system.

[0129] Secondly, at the category level, temporary land use is divided into temporary building land, temporary facility land, temporary activity land, and temporary storage land. Temporary building land mainly includes prefabricated houses, container houses, and temporary sheds; temporary facility land includes temporary roads, bridges, and pipelines; temporary activity land includes exhibition venues and market areas; and temporary storage land includes material storage yards and temporary waste storage areas. Preferably, based on physical characteristics and spatial form, clear discrimination criteria are defined for each category, such as temporary building land typically having a regular shape, clear boundaries, and roof features.

[0130] Third, for the functional layer, temporary land use is divided into production functions, living functions, service functions, and public functions. Production functions mainly refer to temporary land used for economic activities such as manufacturing and resource extraction; living functions include land for temporary accommodation and dining to meet basic living needs; service functions involve temporary venues for commercial, office, and educational services; and public functions include temporary land for public services such as emergency rescue and public health. Preferably, based on the characteristics of land use activities and service recipients, a distinction rule for the functional layer is established, such as differentiating different functional types through indicators like pedestrian density, activity time patterns, and service facility configuration.

[0131] Fourth, the usage layer is further subdivided according to specific application scenarios, including emergency relief and disaster relief, engineering construction, commercial exhibitions and sales, and cultural activities. Emergency relief and disaster relief includes disaster relief sites and temporary medical points; engineering construction includes construction sites and road construction; commercial exhibitions and sales include temporary markets and exhibition venues; and cultural activities include performance venues and celebration venues. Preferably, the usage layer classification is updated regularly in conjunction with urban planning needs and management practices to adapt to changes in urban development and the emergence of new uses.

[0132] Finally, based on weakly supervised learning, a classification model is trained using a small number of labeled samples to achieve automatic classification of large-scale temporary land use. Weakly supervised learning employs an active learning strategy, prioritizing the selection of the most informative and representative samples for labeling to maximize labeling efficiency. Preferably, an uncertainty-based sample selection strategy is used, selecting the samples with the lowest model prediction confidence for labeling, and iteratively optimizing the classification model.

[0133] Active learning sample selection strategies can be implemented using information entropy:

[0134]

[0135] Where H(x) represents the prediction uncertainty of sample x, and p(y) i |x) indicates that sample x belongs to category y. i The predicted probability is given by C, where C represents the total number of categories. Higher information entropy indicates greater prediction uncertainty, making the sample more worthy of labeling.

[0136] Through the steps described above, this embodiment constructs a complete hierarchical classification system for temporary land use, achieving multi-level classification from macro-categories to specific uses. This classification system not only considers the physical characteristics and functional attributes of temporary land use but also incorporates urban planning and management needs, providing a scientific framework for the systematic management of temporary land use. The introduction of weakly supervised learning methods significantly reduces the annotation cost of training the classification model and improves the automation and coverage of the classification.

[0137] Example 6: Establishment of a Change-Sensitive Time Series Model

[0138] In this embodiment, the specific steps for establishing a change-sensitive time series model include:

[0139] First, time window parameters are constructed based on the planning cycle and expected duration of temporary land use. These parameters include the observation start time, observation end time, observation frequency, and key time points, used to define the scope and accuracy of the time series analysis. Preferably, differentiated time window parameters are set according to the characteristics of different types of temporary land use; for example, construction sites typically require longer observation cycles and more frequent observation frequencies, while temporary markets need to capture periodic change patterns.

[0140] Secondly, a change-sensitive loss function tailored to the characteristics of temporary land use is designed to improve the detection sensitivity for changes in small targets and sparsely distributed temporary buildings. This change-sensitive loss function combines focal loss and boundary-sensitive loss, assigning higher weights to difficult-to-distinguish samples and boundary areas. Preferably, the parameters in the loss function are adaptively adjusted according to the data distribution characteristics to ensure good detection capability for changes in temporary land use of different scales and types.

[0141] The mathematical expression for the change-sensitive loss function is:

[0142]

[0143] Where, p t α represents the model's predicted probability for the category of change; α and γ are adjustment parameters for the focus loss (α is typically in the range of 0.25-0.75, and γ is typically 2-5, used to adjust the weight ratio of easy and difficult samples); β is the weight coefficient of the boundary-sensitive term (typically in the range of 0.1-0.5); Ω b y represents the set of pixels representing the boundary region. i and Let represent the true label and predicted value of pixel i, respectively, and ||·||2 represent the L2 norm. The first term focuses on hard-to-classify samples, while the second term emphasizes boundary accuracy.

[0144] Third, a bidirectional temporal encoder is constructed to encode and analyze the diachronic change sequence of temporary land use. Based on a bidirectional long short-term memory network (BiLSTM) and an attention mechanism, the bidirectional temporal encoder can simultaneously capture the forward and backward dependencies of the temporal data. Preferably, the hidden state dimension of the temporal encoder is set to 256, the input feature dimension is consistent with the multimodal feature dimension, and the time step is determined according to the observation frequency.

[0145] The calculation process of a bidirectional timing encoder can be represented as follows:

[0146]

[0147] in, and These represent the hidden states of the forward and backward LSTMs, respectively (the forward LSTM processes from the beginning of the sequence backwards, and the backward LSTM processes from the end of the sequence forwards). This represents the hidden state of the feedforward LSTM at time step t-1. This represents the hidden state of the backward LSTM at time step t+1, x. t LSTM represents the input features at time step t. fand LSTM b Let h represent the computation functions of the forward and backward LSTM units, respectively. t This represents the bidirectional hidden state after connection (usually a concatenation of hidden states in two directions), c t Let α represent the context vector generated through the attention mechanism. t ,i represents the attention weight of time step t to time step i, and T represents the total number of time steps. This indicates a vector concatenation operation.

[0148] Fourth, an urban planning rule base is introduced to constrain anomaly detection with domain knowledge, reducing the false positive rate. The planning rule base includes rules regarding the legal duration of temporary land use, permissible changes, and conversion conditions. Through a combination of rule-based reasoning and machine learning, the model's initial judgments are verified and constrained. Preferably, the rule base employs a knowledge representation method combining ontology and rules, supporting the expression and reasoning of complex rules while possessing the flexibility to update and expand.

[0149] Finally, a multi-evidence fusion method based on evidence theory is employed to quantify the reliability of anomaly detection and generate a confidence score. The evidence theory framework integrates multi-source evidence, including model predictions, rule-based judgments, and historical data, to generate more reliable anomaly detection results. Preferably, Dempster-Shafer evidence theory is used for multi-source evidence fusion, quantifying uncertainty through quality functions, confidence functions, and likelihood functions, thus providing a reliable basis for decision-making.

[0150] The basic formula for multi-evidence fusion is:

[0151]

[0152] Where, m 1,2 (A) represents the basic probability assignment of evidence A after fusion, m1(B) and m2(C) represent the basic probability assignments of the two evidence sources to sets B and C, respectively, and K represents the degree of conflict, defined as... B∩C=A means that the intersection of set B and set C equals set A. This indicates that the intersection of set B and set C is an empty set. This formula can be used to achieve effective fusion of evidence from multiple sources.

[0153] Through the steps described above, this embodiment establishes a change-sensitive time-series model, achieving accurate modeling of temporary land use changes and sensitive detection of anomaly patterns. This model comprehensively utilizes deep learning, attention mechanisms, and knowledge constraints, enabling it not only to detect abnormal temporary buildings that have not been demolished after the planning period, but also to provide reliable confidence assessments, offering a scientific basis for subsequent decision-making.

[0154] Example 7: Temporary Land Use Detection Based on Abnormal Change Patterns

[0155] In this embodiment, as Figure 6 As shown, the specific steps for detecting and identifying temporary land use areas with anomalous change patterns include:

[0156] First, the temporal variation curves of the temporary land use area are fitted to detect outliers that deviate from the expected change pattern. The temporal variation curves are constructed based on multi-temporal remote sensing imagery and ground observation data, reflecting the entire process of temporary land use from its establishment to its demolition or conversion. Preferably, a combination of multinomial regression and seasonal decomposition is used to model the trend component and periodic component respectively, improving the fitting ability to complex change patterns.

[0157] The fitting model for the time-series variation curve can be expressed as:

[0158] Y(t)=T(t)+S(t)+R(t),

[0159] Where Y(t) represents the observed value at time t, T(t) represents the trend component (usually represented by a multinomial function), S(t) represents the seasonal component (usually represented by a Fourier series or seasonal dummy variable), and R(t) represents the residual component (including random fluctuations and anomalous changes). The trend component is obtained through multinomial regression, and the seasonal component is obtained through Fourier transform or seasonal exponential decomposition. Outliers are mainly detected through the statistical properties of the residual component R(t).

[0160] Secondly, anomaly scores are calculated based on the rate of change and the magnitude of change, and temporary land use areas exceeding a preset threshold are marked as potential anomalies. The rate of change refers to the speed of change within a specific time window, and the magnitude of change refers to the difference between the observed value and the expected value. Preferably, the anomaly score uses a combination of Z-score and corrected extreme value analysis to adapt to the change characteristics of different types of temporary land use and reduce the false alarm rate.

[0161] The formula for calculating outlier scores is:

[0162]

[0163] Among them, S anomaly Z(t) represents the outlier score at time t, and Z(·) represents the Z-score standardization (standardizing the data to a distribution with a mean of 0 and a standard deviation of 1). Represents the rate of change (which can be approximated by difference). This represents the difference between the observed and predicted values. ω1 and ω2 are weighting coefficients (satisfying ω1+ω2=1, used to balance the importance of the rate of change and the magnitude of change).

[0164] Third, by combining urban planning cycles and land use conversion rules, the authenticity of potential abnormal areas can be verified. Planning cycles provide official evidence of the expected duration of temporary land use, while land use conversion rules define the target types and conditions under which temporary land use can be converted. Preferably, a planning knowledge graph can be constructed to integrate rules scattered across various planning documents into structured knowledge, enabling efficient rule matching and reasoning.

[0165] Fourth, the identified abnormal temporary land use areas are classified into different types, including those not demolished on schedule, those undergoing functional changes, and those undergoing structural reinforcement. Those not demolished on schedule refer to temporary land that has not been demolished beyond the planned period; those undergoing functional changes refer to temporary land whose use has changed; and those undergoing structural reinforcement refer to temporary land that is gradually being converted from a temporary structure to a permanent structure. Preferably, based on abnormal characteristics and change patterns, a hierarchical clustering method is used to automatically discover abnormal types, improving the adaptability and completeness of the abnormal classification.

[0166] Finally, a spatial distribution heatmap of anomalous temporary land use is generated, visually demonstrating the clustering of anomalous areas. Based on kernel density estimation, the spatial distribution heatmap transforms discrete anomaly points into a continuous density surface, reflecting the spatial clustering characteristics of the anomaly. Preferably, the kernel function and bandwidth parameters of the heatmap are adaptively adjusted according to city size and anomaly distribution characteristics to ensure the heatmap's expressive effect and analytical value.

[0167] The formula for calculating kernel density estimation is:

[0168]

[0169] Where f(x) represents the density estimate at location x, n represents the number of outliers, h represents the bandwidth parameter (controlling the smoothing degree, usually selected through cross-validation to find the optimal value), K(·) represents the kernel function, and x i This represents the position of the i-th outlier. Commonly used kernel functions include the Gaussian kernel and the Epanechnikov kernel.

[0170] Through the steps described above, this embodiment achieves accurate detection and classification of temporary land use patterns exhibiting abnormal changes. This method can not only identify illegal temporary land use that has not been demolished on schedule, but also discover potential transformation trends such as functional shifts and structural reinforcement, providing comprehensive monitoring tools and early warning mechanisms for urban planning and management. Simultaneously, the spatial distribution heat map visually displays the spatial distribution patterns of abnormal phenomena, helping to identify systemic problems and regional characteristics.

[0171] Example 8: Construction of a network diagram model linking temporary land use with urban functions

[0172] In this embodiment, the specific steps for constructing the association graph model between temporary land use and urban functional networks include:

[0173] First, a heterogeneous spatial relationship diagram is constructed based on the spatial relationship between temporary land use and surrounding permanent buildings. This diagram represents temporary land use and permanent buildings as different types of nodes, and spatial relationships as edges, forming a complex network structure with multiple types of nodes and multiple relationship edges. Preferably, spatial relationships include various types such as distance relationships, directional relationships, and topological relationships, comprehensively reflecting the spatial interaction patterns between temporary land use and the surrounding environment.

[0174] The formal definition of a heterogeneous spatial relationship graph is:

[0175] G = (V T ∪V P ,E,R,φ,ψ),

[0176] Where G represents a heterogeneous graph, V T V represents the set of temporary land use nodes. P Let E represent the set of permanent building nodes, R represent the set of edges, and φ represent the set of relation types (such as distance relations, functional relations, etc.). T ∪V P →A represents the mapping from a node to an attribute (A is the set of attributes), and ψ:E→R represents the mapping from an edge to a relation type. Next, the association strength between the temporary land use and surrounding urban functional nodes is calculated, including indicators such as spatial proximity, functional complementarity, and frequency of pedestrian interaction. Spatial proximity is calculated based on Euclidean distance or network distance, functional complementarity measures the degree of functional cooperation, and the frequency of pedestrian interaction is estimated based on movement trajectory data. Preferably, the association strength calculation adopts a multi-indicator comprehensive evaluation method, determining the weight of each indicator through the entropy weight method to generate a comprehensive association strength index.

[0177] The formula for calculating the correlation strength is:

[0178]

[0179] Wherein, S(v i ,v j ) represents node v i and v j The overall correlation strength between them, s k (v i ,v j ) represents the correlation strength of the k-th indicator (such as spatial proximity, functional complementarity, etc.), w k This represents the weight of the k-th indicator (satisfying...) K represents the total number of indicators. Weight w k The index values ​​are calculated using the entropy weight method and then normalized to ensure comparability.

[0180] Third, a multi-layer graph convolutional network is employed to extract the structured and functional features of temporary land use nodes. Through a message-passing mechanism, the multi-layer graph convolutional network diffuses and aggregates the local information of nodes layer by layer, forming node representations rich in contextual information. Preferably, considering the characteristics of heterogeneous graphs, relation-aware graph convolution operations are used, employing different parameter matrices for different types of relations to improve the accuracy of feature extraction.

[0181] Relation-aware graph convolution operations can be represented as:

[0182]

[0183] in, This represents the feature representation (vector form) of a node at layer l. This represents the feature representation of a node at level l+1, where r∈R indicates that relation type r belongs to the set of relation types R. This represents the set of neighboring nodes that are connected to node i through relation r. Represents a set The number of elements (used for normalizing neighbor information), Let r represent the weight matrix (learnable parameters) of relation r at layer l. The weight matrix represents the self-loop (used to preserve the node's own information), and σ represents the activation function (such as ReLU, tan h, etc.).

[0184] Fourth, spectral clustering algorithms are used to discover temporary land use communities with similar functional patterns and evolution trends. Spectral clustering, based on the eigenvectors of the graph Laplacian matrix, effectively captures community characteristics within the network structure. Preferably, the number of clusters is adaptively determined, and the optimal clustering scheme is evaluated through eigenvalue analysis and silhouette coefficients to discover temporary land use functional communities with inherent connections.

[0185] The core step in spectral clustering is to calculate the graph Laplacian matrix: L = DA.

[0186] Where L represents the graph Laplacian matrix, and D represents the degree matrix (diagonal elements D) ii Let A be the degree of node i (i.e., the number of edges connected to node i), and let A be the adjacency matrix (elements A...). ij The value indicates whether there is an edge connecting node i and node j (1 if there is, 0 if there isn't). By calculating the eigenvectors of L and applying traditional clustering algorithms such as K-means in the feature space, the community structure in the graph can be discovered.

[0187] Finally, the location and importance of abnormal temporary land use within the urban functional network are analyzed, and its preservation or redevelopment value is assessed. Locational importance is measured using centrality indicators, including degree centrality, betweenness centrality, and eigenvector centrality; functional importance considers its role and contribution in meeting urban functional needs. Preferably, a comprehensive evaluation system is constructed by combining socio-economic indicators and spatial planning objectives to scientifically assess the priority of disposal and redevelopment value of abnormal temporary land use.

[0188] Through the steps described above, this embodiment constructs a graph model relating temporary land use to urban functional networks, achieving a methodological leap from point-based analysis to network-based understanding. This model not only reveals the complex interaction between temporary land use and urban functional systems but also identifies clusters of temporary land use with similar functional patterns, providing a systematic analytical perspective and scientific basis for urban planning.

[0189] Example 9: Generation of Differentiated Urban Planning Optimization Suggestions

[0190] In this embodiment, the specific steps for generating differentiated urban planning optimization suggestions include:

[0191] First, a multi-objective evaluation index system should be constructed, encompassing economic benefits, social value, and environmental impact. Economic benefit indicators include land use efficiency, industrial driving force, and return on investment; social value indicators include service coverage, improvement in people's livelihoods, and equitable sharing; and environmental impact indicators include ecological impact, resource consumption, and pollution emissions. Preferably, differentiated indicator weights should be set according to the priorities of urban planning and the characteristics of each city, reflecting the development needs and unique positioning of different cities.

[0192] The comprehensive score calculation method for multi-objective evaluation is as follows:

[0193]

[0194] Where S represents the overall score, w i Represents the weight of the i-th indicator (satisfying) ), x i Min represents the actual value of the i-th indicator. i and max i Let represent the minimum and maximum values ​​of the i-th indicator, respectively, and N represent the total number of indicators. This calculation method normalizes each indicator to the [0,1] interval to ensure the comparability of indicators with different dimensions. Secondly, based on cellular automata and system dynamics models, urban evolution scenarios under different planning strategies are simulated. Cellular automata simulate the evolution of urban spatial morphology, while system dynamics models simulate the complex feedback relationships between various elements of the urban system. Preferably, the two models are coupled into a unified simulation framework to achieve collaborative simulation of spatial dynamics and system dynamics, improving the accuracy and completeness of scenario prediction.

[0195] The state transition rules of a cellular automaton can be expressed as:

[0196]

[0197] in, This represents the state of location (i,j) at time t (e.g., land use type). Let f represent the set of neighborhood states of location (i,j) at time t (usually Moore's neighborhood or von Neumann's neighborhood), and let f represent the state transition function (which determines the state at the next time step based on the neighborhood states and the transition rules). By defining reasonable transition rules, the evolution process and spatial diffusion pattern of temporary land use can be simulated.

[0198] Third, based on the characteristics of abnormal temporary land use and urban functional needs, personalized planning recommendations are generated, including demolition and reconstruction, functional upgrades, and structural modifications. Demolition and reconstruction are suitable for temporary land use with no preservation value and that does not conform to planning regulations; functional upgrades are suitable for temporary land use that contributes to improving urban functions; and structural modifications are suitable for temporary land use that requires optimization and improvement. Preferably, a case-based reasoning approach is adopted to retrieve similar scenarios from a historical planning case database, extract experiences and lessons learned, and assist in formulating targeted planning recommendations.

[0199] The process of generating personalized planning suggestions can be represented as follows:

[0200]

[0201] Where R represents the optimal planning suggestion. Represents the set of optional suggestions, sim(C r C t ) represents Case C r With Target Case C t The similarity is usually expressed as cosine similarity or the reciprocal of Euclidean distance. E(r) represents the historical validity of the suggestion r (success rate based on historical cases). F(r,G) represents the coordination between the suggestion r and the urban functional network G (to assess the impact of the suggestion on the urban functional network after implementation). α, β and γ are weighting coefficients (satisfying α+β+γ=1, used to balance the importance of similarity, validity and coordination).

[0202] Fourth, provide visualization analysis tools to intuitively display the expected effects and impact assessments of the planning recommendations. These tools include various formats such as 2D floor plans, 3D scene diagrams, indicator radar charts, and impact relationship diagrams, comprehensively showcasing the spatial layout, morphological characteristics, performance indicators, and systemic impacts of the planning recommendations. Preferably, immersive interactive technology is employed to support planners in browsing, comparing, and adjusting different plans, enhancing the intuitiveness and interactivity of planning decisions.

[0203] Finally, a collaborative decision-making mechanism involving multiple parties is designed to integrate feedback from planners, citizens, and developers to optimize the planning scheme. This mechanism establishes communication channels and consultation spaces among diverse stakeholders through a combination of online platforms and offline workshops. Preferably, a combination of the Analytic Hierarchy Process (AHP) and the Delphi method is employed to scientifically collect and integrate opinions from all parties, resulting in a planning scheme with broad consensus.

[0204] Through the steps described above, this embodiment generates differentiated urban planning optimization suggestions. This method transforms the anomaly detection results and functional correlation analysis results of temporary land use into specific planning actions, forming a complete closed loop from data analysis to decision support. The introduction of mechanisms such as multi-objective evaluation, scenario simulation, personalized suggestions, and collaborative decision-making ensures the scientific rigor, forward-looking nature, and feasibility of the planning suggestions, effectively supporting the efficient allocation and management of urban land resources.

[0205] Example 10: GIS-based Urban Land Planning and Analysis System

[0206] like Figure 7 As shown, this embodiment of the invention also provides a GIS-based urban land planning and analysis system, including a processor and a memory; wherein, the memory stores a computer program, and when the computer program is executed by the processor, it implements the GIS-based urban land planning and analysis method as described in embodiments 1-9.

[0207] The system mainly includes the following functional modules:

[0208] Multi-source data acquisition module 1 is responsible for acquiring multi-source heterogeneous spatiotemporal data from remote sensing satellite platforms, urban planning databases, mobile terminal devices, and social media platforms, and performing quality assessment and preprocessing.

[0209] The Spatiotemporal Cube Construction Module 2 is responsible for spatiotemporal alignment of multi-source heterogeneous data, constructing a four-dimensional spatiotemporal data structure, and realizing unified management of multi-scale spatial information and time series interpolation processing.

[0210] The adaptive feature extraction module 3 is responsible for building a multi-scale feature extraction network, fusing multi-dimensional features, optimizing feature weight allocation, and achieving accurate identification and attribute labeling of temporary land use areas.

[0211] The hierarchical classification management module 4 is responsible for establishing a three-level classification framework, realizing the classification of temporary land use by category, function and use, and supporting large-scale automatic classification through weakly supervised learning methods.

[0212] The time-series change analysis module 5 is responsible for constructing time window parameters, designing change-sensitive loss functions, and realizing the modeling of temporary land use change processes and the detection of abnormal patterns.

[0213] The anomaly identification and assessment module 6 is responsible for fitting and detecting anomalies in the temporal changes of temporary land use, verifying the authenticity of anomalies in conjunction with planning rules, and generating anomaly classification and spatial distribution heatmaps.

[0214] The Relationship Network Analysis Module 7 is responsible for constructing a graph model of the relationship between temporary land use and urban functions, calculating the relationship strength, extracting structured features, discovering functional clusters, and assessing the locational importance of abnormal temporary land use.

[0215] The planning suggestion generation module 8 is responsible for constructing a multi-objective evaluation index system, simulating urban evolution scenarios under different planning strategies, generating personalized planning suggestions, and providing visualization analysis tools and collaborative decision-making mechanisms.

[0216] The data storage management module 9 is responsible for managing the system's data resources, including raw data, intermediate processing results, and analysis results, and supports efficient data retrieval and updates.

[0217] User interface module 10 is responsible for providing a user-friendly human-computer interaction interface, supporting data visualization, analysis operations and result display, and improving system usability and user experience.

[0218] The various functional modules interact through standardized data interfaces and service protocols, forming a complete data processing and analysis pipeline. The system architecture adopts a microservice design, allowing each module to be deployed and expanded independently to meet the application scenarios of cities of different sizes and with different analytical needs.

[0219] In its implementation, the system can be deployed on a cloud platform or a local server, supporting multi-terminal access via web and mobile clients. The system's core algorithms and models utilize GPU acceleration for efficient computation, supporting parallel processing and real-time analysis of large-scale data.

[0220] The system provided in this embodiment is based on a microservices architecture and cloud-native technologies, achieving a highly modular, scalable, and easily maintainable design. It offers urban planning and management departments a powerful tool for temporary land use analysis and decision support. The system not only meets routine temporary land use monitoring and management needs but also supports forward-looking research and innovative practices in urban planning.

[0221] To further illustrate the practical application effects of the present invention, an actual case is provided below.

[0222] This case study selects a rapidly developing new district as its research object, focusing on the analysis and planning optimization of temporary building land use within the area. The study area covers approximately 50 square kilometers and includes various types of temporary building land use, such as construction sites, temporary markets, and temporary event venues.

[0223] First, the system acquired spatiotemporal data for the region from multiple data sources, including high-resolution satellite imagery from the past five years (0.5-meter resolution), historical planning documents from an urban planning database, ground-based photographs provided by over 300 volunteers, and geotagged data from social media platforms. Data quality assessment results showed that the overall quality score for the satellite imagery was 0.92, the effective coverage of the ground photographs was 78%, and the spatiotemporal uniformity of the social media data was 0.65.

[0224] Secondly, the system constructed a multi-scale spatiotemporal cube model of the region, achieving unified management of data from different sources. The average accuracy of spatiotemporal alignment reached sub-meter level, the multi-scale spatial representation supports five scales from 0.5 meters to 100 meters, and the interpolation accuracy of time series data exceeded 95%.

[0225] Third, the adaptive feature extraction mechanism successfully identified 327 temporary land use areas within the region, covering a total area of ​​approximately 3.2 square kilometers. The identification accuracy reached 92.3% through manual verification, significantly higher than the 68.7% of traditional single-data source methods. The system categorizes these temporary land use areas into 4 category layers, 4 function layers, and 12 usage layers, forming a complete classification system.

[0226] Fourth, the change-sensitive time-series model analysis revealed that among the 327 temporary land uses, 42 exhibited abnormal change patterns. Of these, 28 were due to failure to demolish on schedule, 9 were due to functional changes, and 5 were due to structural reinforcement. The anomaly detection accuracy was 88.1%, the recall rate was 91.3%, and the F1 score was 89.7%.

[0227] Fifth, network analysis revealed the complex connections between these 42 unusual temporary land uses and the functions of the surrounding cities. A heterogeneous relationship graph containing over 2,500 nodes and 15,000 edges was constructed. Node features were extracted using a multi-layer graph convolutional network, identifying six temporary land use clusters with similar functional patterns. Analysis showed that temporary commercial clusters located near transportation hubs possessed the highest functional complementarity and preservation value.

[0228] Finally, based on multi-objective evaluation and scenario simulation, the system generated differentiated planning recommendations for these 42 abnormal temporary land uses: demolition and reconstruction for 12 sites, functional upgrades for 18 sites, structural modifications for 8 sites, and overall preservation for 4 sites. These recommendations were discussed and optimized through a collaborative decision-making mechanism involving multiple parties, ultimately forming a planning scheme with a broad base of acceptance.

[0229] The results of this case study demonstrate that the method and system provided by this invention can effectively support the analysis and planning decisions regarding temporary urban land use. Through a complete process involving multi-source data fusion, feature extraction, anomaly detection, correlation analysis, and planning suggestion generation, the comprehensiveness, accuracy, and practicality of temporary land use analysis are significantly improved, providing a scientific basis and decision support for urban planning and management.

[0230] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A GIS-based urban land planning analysis method, characterized in that, include: Acquire multi-source heterogeneous spatiotemporal data, including remote sensing image data, historical planning data, mobile terminal data, and social media geotagged data; Based on the acquired multi-source heterogeneous spatiotemporal data, a unified multi-scale spatiotemporal cube model is constructed, and standardized multimodal urban temporary land use characterization data is generated. Based on the standardized multimodal urban temporary land use characterization data, an adaptive feature extraction mechanism is used to identify urban temporary land use areas and a hierarchical classification system for temporary land use is constructed. Based on the hierarchical classification system and historical time series data of the temporary land use, a change-sensitive time series model is established to detect and identify temporary land use areas with abnormal change patterns. Based on the identified temporary land use areas with abnormal change patterns, a correlation diagram model between temporary land use and urban functional networks is constructed to analyze the interaction between abnormal temporary land use and urban functional systems. Based on the aforementioned interaction relationships and multi-objective evaluation indicators, differentiated urban planning optimization suggestions are generated to support the efficient allocation and management of urban land resources.

2. The GIS-based urban land planning analysis method according to claim 1, characterized in that, The acquisition of multi-source heterogeneous spatiotemporal data includes: Acquire optical and radar remote sensing image data of the target urban area from a remote sensing satellite platform; Retrieve historical planning data and land use information for the target urban area from the urban planning database; Collect ground-based real-view data and location information of the target urban area using mobile terminal devices; Extract geotagged data and user activity information related to target city areas from social media platforms; The quality of the acquired data is assessed, and a data reliability scoring system is constructed based on information entropy and texture features to determine the weight coefficients of each data source.

3. The GIS-based urban land planning analysis method according to claim 1, characterized in that, The construction of a unified multi-scale spatiotemporal cube model includes: Spatiotemporal alignment of multi-source heterogeneous data is performed based on feature point matching algorithm to ensure consistency of data from different sources in time and space dimensions; Construct a four-dimensional spatiotemporal data structure that integrates spatial coordinates (x, y, z) and the time dimension (t) into a unified data representation model; Scale transformation is performed on spatial data of different resolutions to achieve unified management of multi-scale spatial information; Interpolation is performed on discontinuous time series data to establish continuous and equally spaced time series data. Perform semantic-level fusion and standardization of multi-source data to ensure semantic consistency of data from different sources.

4. The GIS-based urban land planning analysis method according to claim 1, characterized in that, The method of identifying temporary urban land use areas using an adaptive feature extraction mechanism includes: A multi-scale feature extraction network is constructed to extract hierarchical features from urban area images; By integrating spectral features, texture features, morphological features, and contextual features, a multidimensional feature representation of temporary land use is constructed. Optimize feature weight allocation based on attention mechanism to highlight key distinguishing features of temporary land use; Accurate boundary extraction of temporary land use areas is achieved through a context-aware segmentation algorithm. The extracted temporary land use areas are labeled with attributes, including area, shape, material, and usage information.

5. The GIS-based urban land planning analysis method according to claim 1, characterized in that, The hierarchical classification system for temporary land use includes: Establish a three-level classification framework, including a category layer, a function layer, and a usage layer; For the category level, temporary land use areas are divided into temporary building land, temporary facility land, temporary activity land, and temporary storage land; For the functional layer, temporary use areas are divided into production functions, living functions, service functions, and public functions; For the usage layer, it is further subdivided into emergency rescue and disaster relief, engineering construction, commercial exhibitions and sales, and cultural activities based on specific application scenarios; Based on weakly supervised learning, a classification model is trained using a small number of labeled samples to achieve automatic classification of large-scale temporary land use.

6. The GIS-based urban land planning analysis method according to claim 1, characterized in that, The establishment of the change-sensitive time series model includes: Based on the planning cycle and expected duration of temporary land use, construct time window parameters; Design a loss function that is sensitive to changes in the characteristics of temporary land use to improve the detection sensitivity of changes in small targets and sparsely distributed temporary buildings; A bidirectional time-series encoder was constructed to encode and analyze the epochal change sequence of temporary land use; By introducing an urban planning rule base, anomaly detection is constrained by domain knowledge, thereby reducing the false judgment rate; A multi-evidence fusion method based on evidence theory is used to quantify the reliability of anomaly detection and generate a confidence score.

7. The GIS-based urban land planning analysis method according to claim 1, characterized in that, The temporary land use areas for detecting and identifying abnormal change patterns include: The temporal variation curves of temporary land use areas are fitted to detect outliers that deviate from the expected change patterns; Anomaly scores are calculated based on the rate of change and the magnitude of change, and temporary land use areas that exceed the preset threshold are marked as potential anomalies. By combining urban planning cycles and land use conversion rules, the authenticity of potential abnormal areas can be verified; The identified abnormal temporary land use areas are classified into three types: those not demolished on schedule, those undergoing functional transformation, and those requiring structural reinforcement. Generate a spatial distribution heat map of abnormal temporary land use to visually demonstrate the clustering of abnormal areas.

8. The GIS-based urban land planning analysis method according to claim 1, characterized in that, The construction of the relationship graph model between temporary land use and urban functional network includes: Based on the spatial relationship between temporary land use and surrounding permanent buildings, construct a heterogeneous spatial relationship diagram; Calculate the correlation strength between temporary land use and surrounding urban functional nodes, including spatial proximity, functional complementarity, and frequency of pedestrian interaction. A multi-layer graph convolutional network is used to extract the structural and functional features of temporary land use nodes; Temporary land use communities with similar functional patterns and evolution trends were discovered based on spectral clustering algorithms; Analyze the location and importance of abnormal temporary land use within the urban functional network, and assess its value for preservation or redevelopment.

9. The GIS-based urban land planning analysis method according to claim 1, characterized in that, The generated differentiated urban planning optimization suggestions include: Construct a multi-objective evaluation index system that includes economic benefits, social value, and environmental impact; Based on cellular automata and system dynamics models, we simulate urban evolution scenarios under different planning strategies. Based on the characteristics of abnormal temporary land use and the needs of urban functions, generate personalized planning suggestions, including demolition and reconstruction, functional upgrades, and structural renovation plans; Provides visualization and analysis tools to intuitively display the expected effects and impact assessments of planning recommendations; Design a collaborative decision-making mechanism involving multiple parties, integrate feedback from planners, citizens, and developers, and optimize planning schemes.

10. A GIS-based urban land planning and analysis system, characterized in that, It includes a processor and a memory; wherein the memory stores a computer program, which, when executed by the processor, implements the GIS-based urban land planning analysis method as described in any one of claims 1 to 9.

Images