A land potential evaluation method, system, device and storage medium for urban planning

Abstract

The present application relates to the technical field of urban planning, and specifically provides a land potential evaluation method, system, device and storage medium for urban planning, which comprises: preprocessing the natural language demand input by a user to obtain a planning demand text, identifying and evaluating the intention, extracting the core target and the limiting condition through a planning language processing model, and generating demand-oriented parameters containing classification labels, core parameters and constraint thresholds; obtaining and preprocessing multi-source planning land spatial data and text data to form a planning data set, and based on the demand-oriented parameters, extracting rigid constraints and elastic guide type evaluation constraint parameters; fusing the related parameters and the spatial data to obtain an evaluation input data set, inputting the planning land potential evaluation model, combining the index system and the weight rule to complete comprehensive quantitative evaluation, and finally outputting the quantitative evaluation score and the potential level label. The present application can improve the automation, standardization level and response efficiency of the evaluation, and provide accurate support for planning decision-making.

Description

Technical Field

[0001] This invention relates to the field of urban planning technology, specifically to a method, system, device, and storage medium for assessing the potential of urban planning land use. Background Technology

[0002] Land use potential assessment is a core prerequisite for urban spatial planning, existing land quality improvement and renewal, and optimal allocation of land resources. It directly affects the scientific nature of planning decisions, their feasibility, and resource utilization efficiency. As my country's new urbanization enters a high-quality development stage, urban planning places higher demands on the accuracy, efficiency, and adaptability of land use assessments. The assessment work not only needs to integrate structured spatial data such as spatial boundaries and development intensity, but also unstructured textual data such as planning-related documents and technical standards. At the same time, it must conform to spatial planning management rules to respond to differentiated assessment needs (such as residential land development intensity assessment, land use suitability assessment within ecological protection boundaries, and analysis of the renewal potential of existing industrial land plots), providing core basis for determining land development intensity, optimizing functional layout, and verifying planning schemes.

[0003] However, traditional land use potential assessment has long been hampered by technical challenges such as difficulty in integrating multi-source heterogeneous data, high professional barriers, low efficiency, and significant subjective bias. The entire assessment process heavily relies on manual operation by professional technicians: it requires manually organizing multi-source data from different channels, calculating and verifying thresholds for various indicators such as rigid compliance and flexible potential, and conducting complex spatial correlation analysis and rule adaptation in conjunction with planning technical standards. This process is not only time-consuming and involves a high proportion of repetitive work, but also prone to inconsistencies in assessment results due to differences in experience and cognitive biases among different assessors, making it difficult to form a unified, traceable, and standardized assessment output. In addition, traditional assessment methods often rely on fixed templates and static parameter configurations. When faced with diverse and dynamic assessment needs, it is necessary to redevelop algorithm logic and adjust parameters, resulting in slow response speed, insufficient adaptability, and inability to meet the needs of rapid iterative planning decisions.

[0004] Against this backdrop, there is an urgent need to develop a land use potential assessment method that is adapted to urban planning business scenarios. Through standardized process design, efficient integration of multi-source data, and intelligent analysis logic, the assessment work can be automated, standardized, and efficient, solving the core problems of traditional assessments such as high dependence on manual labor, poor consistency of results, and insufficient adaptability to needs. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, the present invention provides a method, system, device and storage medium for assessing the potential of urban planning land use, in order to solve the problems in the prior art.

[0006] One embodiment of the present invention provides a method for assessing the potential of urban planning land use, comprising the following steps: S10. In response to the user's input of natural language requirements, the natural language requirements are preprocessed to obtain the planning requirement text; S20. Input the planning requirement text into the planning language processing model, identify the user's planning evaluation intention type and extract the core evaluation objectives and additional constraints, integrate the planning evaluation intention type, core evaluation objectives and additional constraints to generate a demand-oriented parameter that includes classification labels, core parameters and constraint thresholds. S30. Obtain multi-source planned land use data including spatial data and text data, and preprocess the multi-source planned land use data to obtain a planning dataset; S40. Guided by demand-oriented parameters, the planning dataset is parsed and extracted using a planning language processing model to generate evaluation constraint parameters associated with the demand-oriented parameters. The evaluation constraint parameters have rigid constraint rules and flexible guidance rules. S50. The demand-oriented parameters, evaluation constraint parameters and spatial data in the planning dataset are fused to obtain the evaluation input dataset; S60. Input the assessment input dataset into the planning land potential assessment model, combine the preset assessment index system and weight allocation rules to conduct a comprehensive quantitative assessment, and output the planning land potential assessment result. The planning land potential assessment result includes at least a quantitative assessment score and a potential level label.

[0007] This application also relates to a system for assessing the potential of urban planning land use, including: The demand response module is used to respond to the natural language demands input by the user, preprocess the natural language demands, and obtain the planning demand text. The requirement extraction module is used to input the planning requirement text into the planning language processing model, identify the user's planning evaluation intention type and extract the core evaluation objectives and additional constraints, integrate the planning evaluation intention type, core evaluation objectives and additional constraints, and generate requirement-oriented parameters containing classification labels, core parameters and constraint thresholds. The data acquisition module is used to acquire multi-source planned land use data, including spatial data and text data, and to preprocess the multi-source planned land use data to obtain a planning dataset. The constraint extraction module is used to parse and extract planning datasets using a planning language processing model, guided by demand-oriented parameters, to generate evaluation constraint parameters associated with the demand-oriented parameters. The evaluation constraint parameters have rigid constraint rules and flexible guidance rules. The data fusion module is used to fuse the demand-oriented parameters, evaluation constraint parameters and spatial data in the planning dataset to obtain the evaluation input dataset; The evaluation output module is used to input the evaluation input dataset into the planning land potential evaluation model, combine it with the preset evaluation index system and weight allocation rules to conduct a comprehensive quantitative evaluation, and output the planning land potential evaluation result. The planning land potential evaluation result includes at least a quantitative evaluation score and a potential level label.

[0008] This application also relates to a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for assessing urban planning land use potential.

[0009] This application also relates to a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for assessing the potential of urban planning land use.

[0010] The urban planning land use potential assessment method, system, equipment, and storage medium provided in the above embodiments have the following beneficial effects: This application enables the assessment of the potential of planned land use through natural language interaction, which lowers the professional threshold for such assessments. It allows for intelligent adaptation of assessment needs to multi-source planning data, addressing the heavy reliance on manual labor in traditional assessments, reducing repetitive work, and shortening the assessment cycle. Furthermore, it unifies rule parsing and quantitative assessment, avoiding subjective human bias and ensuring the consistency and traceability of assessment results. Moreover, it flexibly adapts to diverse and dynamic assessment needs without frequent algorithm parameter adjustments, significantly improving the automation, standardization, and response efficiency of assessments, and providing precise support for planning decisions. Attached Figure Description

[0011] Figure 1 A flowchart of a method for assessing urban planning land use potential provided in an embodiment of the present invention; Figure 2 This is a schematic block diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0012] The technical solutions in the embodiments of the present invention will now be clearly and completely described in conjunction with the accompanying drawings.

[0013] Reference Figure 1 One embodiment of the present invention provides a method for assessing the potential of urban planning land use, comprising the following steps: S10. In response to the user's input of natural language requirements, the natural language requirements are preprocessed to obtain the planning requirement text.

[0014] This step aims to respond to user requests for land use potential assessment submitted via an interactive interface in natural language. Examples include colloquial and non-standardized expressions such as, "Could you help me assess the development potential of approximately 10 hectares of residential land in a certain district of a certain city, preferably meeting ecological requirements?" The process involves standardized preprocessing of user-input natural language requests, including: removing redundant information by eliminating irrelevant modifiers, phrases, and repetitive expressions such as "help me," "approximately," "how," and "best," which lack practical assessment significance; standardizing planning terminology by unifying colloquial and heterogeneous expressions into standard planning terminology, such as unifying "residential land" and "residential-type land" as "urban residential land (R category); and standardizing sentence structure by organizing colloquial and fragmented statements into standardized sentences of "assessment + object + goal + constraint," eliminating disordered word order and unclear logic, resulting in a semantically clear, terminologically standardized, and grammatically consistent planning request text.

[0015] S20. Input the planning requirement text into the planning language processing model, identify the user's planning evaluation intention type, extract the core evaluation objectives and additional constraints, and integrate the planning evaluation intention type, core evaluation objectives, and additional constraints to generate demand-oriented parameters containing classification labels, core parameters, and constraint thresholds. The specific steps are as follows: S211. Input the standardized requirement text into the planning language processing model, extract the deep semantic features in the standardized requirement text through the semantic encoding layer of the model, and generate a semantic requirement feature vector.

[0016] Specifically, the obtained planning requirement text (such as "Assess the development potential of 10 hectares of urban residential land in a certain district of a certain city, which must comply with ecological constraints") is input into a pre-built and trained planning language processing model. The semantic encoding layer of the model captures the literal semantics of the planning requirement text through a Transformer encoder structure and mines the implicit logical connections in the text (such as the object-scope association between 10 hectares and urban residential land, and the target-constraint association between development potential and ecological constraints), to obtain deep semantic features. Based on the extracted deep semantic features, a fixed-dimensional semantic requirement feature vector is generated. The vector contains a quantitative expression of key information such as land parcel scope, land use type, assessment target, and core constraints (for example, dimensions 3-5 of the vector correspond to the coordinate features of the assessment scope, and dimensions 6-8 correspond to the coding features of the application land type), realizing the transformation from textual expression to quantitative features.

[0017] S212. Based on the pre-set planning domain knowledge graph, the semantic requirement feature vector is aligned and semantically enhanced through the planning knowledge fusion layer of the model, and planning-adaptive semantic features are output. Specifically, a planning domain knowledge graph covering the core business of territorial spatial planning is pre-constructed (such as a knowledge graph containing land use classification systems, constraint indicator association rules, and logical relationships of assessment elements, where nodes are urban residential land, plot ratio, ecological protection boundaries, etc., and edges are logical relationships such as requirements and related indicators). The planning knowledge fusion layer of the model associates and matches semantic requirement feature vectors with this knowledge graph: first, a feature alignment operation is performed to map non-standardized semantic features in the vectors to standard nodes in the knowledge graph (such as aligning ecological constraint features to related nodes such as ecological protection boundary constraints and green space ratio thresholds in the knowledge graph); then, a semantic enhancement operation is performed to supplement the professional information implicit in the requirements based on the logical relationships of the knowledge graph (such as supplementing the implicit semantics of the need for supporting kindergartens and primary schools for residential land development potential assessment based on the association between urban residential land and educational facilities); and finally, a planning-adaptive semantic feature that integrates professional knowledge, is more semantically complete, and has stronger adaptability is output.

[0018] S213. Based on planning-adaptive semantic features, the model's intent parsing output layer is used to process the data to identify the user's planning evaluation intent type and extract the core evaluation objectives and additional constraints; among which, the additional constraints include at least the evaluation scope, priority requirements, and compliance constraints. Specifically, based on the obtained planning-adaptive semantic features, the model's intent parsing output layer completes two core operations through a multi-classifier and a feature extractor: First, it identifies the type of planning assessment intent by matching a pre-defined intent classification label library (such as residential land development potential assessment, industrial land update adaptability analysis, ecological protection zone land use compliance verification, etc.) to determine the core business scenario of the user's needs (such as determining the intent type of assessing the development potential of 10 hectares of urban residential land in a certain district of a certain city as residential land development potential assessment); Second, it extracts core elements by using named entity recognition (NER) and relation extraction (RE) algorithms to extract core assessment objectives (such as quantitative assessment of residential land development potential) and additional constraints. The additional constraints mainly include three core dimensions: assessment scope (such as a certain district of a certain city, with an area of ​​10 hectares), priority requirements (such as ecological protection taking precedence over development intensity), and compliance constraints (such as plot ratio ≤ 2.5 and green space ratio ≥ 30%).

[0019] S214. Structure and integrate the planning assessment intent type, core assessment objectives and additional constraints to generate demand-oriented parameters that include classification labels, core parameters and constraint thresholds.

[0020] Specifically, based on the parameterization requirements of planning assessment business, the extracted planning assessment intent types, core assessment objectives, and additional constraints are structurally mapped and integrated: the planning assessment intent types are converted into classification labels, with the format of [Assessment Task Name]-[Land Use Type Code]-[Assessment Dimension], such as Residential Land Development Potential Assessment-R-Potential; the core assessment objectives and key attributes are converted into core parameters, such as Assessment Object: 10 hectares of urban residential land in a certain district of a certain city, Assessment Objective: Quantitative score of development potential; the quantitative constraints in the additional constraints are converted into constraint thresholds, such as plot ratio ≤ 2.5, green space ratio ≥ 30%, no construction within the ecological protection boundary; after field standardization and logical verification (such as checking whether there is a spatial conflict between the assessment scope and compliance constraints), standardized demand-oriented parameters containing three major modules: classification labels, core parameters, and constraint thresholds are generated.

[0021] It should be noted that the planning language processing model is a pre-built and trained model used to automate and professionally parse users' natural language needs. It transforms unstructured natural language expressions into standardized, machine-recognizable demand-oriented parameters, addressing the pain points of high demand communication costs, high professional barriers, and low efficiency of manual conversion in traditional assessments. This provides clear demand basis for directional constraint extraction and assessment. Its construction and training steps are as follows: S201. Using a basic language model as the initial model, construct the network structure of the initial model. The network structure includes a semantic encoding layer, a planning knowledge fusion layer, and an intent parsing output layer. Input the pre-constructed planning training set into the initial model for incremental pre-training, so that the initial model learns the basic semantic features of the planning domain.

[0022] Specifically, mature open-source basic language models in the industry are selected as the initial models (such as the Llama series, BERT series, ChatGLM series, and other general-purpose large language models). These models already possess general natural language encoding and basic text processing capabilities. Based on this, a three-layer network structure adapted to the professional processing needs of planning is constructed, including a semantic encoding layer, a planning knowledge fusion layer, and an intent parsing output layer. Among them, the semantic encoding layer is used to perform deep semantic vector encoding on the input text, extracting the basic semantic and sentence structure features of the text, which is the basic unit for the model to achieve semantic understanding. The planning knowledge fusion layer is used to align and fuse the encoded semantic features with professional knowledge, constraint rules, and standard paradigms in the planning domain, thereby enhancing the adaptability of professional semantics. The intent parsing output layer undertakes the final intent determination, element extraction, and structured output functions, which is the output unit for the model to connect with business needs. The three-layer architecture works together to form the overall processing framework of the planning language processing model.

[0023] The pre-constructed and standardized planning training set is input into the initial model in a preset batch to perform incremental pre-training in the planning domain. During the training process, the model learns the terminology, logic and expression paradigms in planning professional texts through the semantic encoding layer, autonomously extracts basic semantic features in the planning domain, weakens general semantic interference, and completes the ability transfer from general language processing to basic semantic cognition in the planning domain, providing an adapted basic model for subsequent knowledge fusion and intent parsing.

[0024] For the pre-constructed planning training set, different types of texts were collected and integrated to construct the set around the core needs of planning land potential assessment and territorial spatial planning business, including: (1) planning constraint texts, such as territorial spatial master plan, control detailed plan, land use zoning constraint requirements, land development constraint rules, etc.; (2) planning spatial texts, such as special plans, area renewal plans, land use layout plans, existing land revitalization plans, etc.; (3) planning technical texts, such as territorial spatial land use classification, planning technical management regulations, development intensity control specifications, public service facility supporting standards, etc.; (4) land use assessment demand texts, such as planning project assessment demand, land use suitability analysis demand, etc.

[0025] It should be noted that all of the above-mentioned texts can be obtained by those skilled in the art through public channels and compliant methods. For example, planning constraint texts and planning spatial texts can be searched and downloaded from public channels such as the official publicity platforms of the natural resources authorities, public websites, and planning results filing systems; planning technical texts are standards and specifications publicly released by the state, industry, and local governments, and can be accessed free of charge and obtained compliantly through the official website of the National Standardization Management Committee, the release platforms of industry authorities, and standard public service platforms; land use assessment requirement texts can be prepared and generated by referring to publicly available planning business scenarios (such as publicly available planning consultation cases and typical requirement scenarios published in industry seminars) according to the conventional expression format of planning assessment, or by using de-identified general requirement case texts to ensure that data acquisition and use are legal and compliant, which is clear to those skilled in the art.

[0026] The collected text data is cleaned to remove duplicate, invalid, and erroneous content, standardize the text format and planning terminology, remove irrelevant and interfering information, and then undergo standardized integration to construct a professional and standardized planning training set.

[0027] S202. Based on the text semantic features in the planning training set and the preset model training objectives, construct a supervised task that includes a planning terminology recognition task, a planning intent classification task, and a land use demand element extraction task.

[0028] Specifically, the semantic features of the texts in the planning training set are analyzed, such as the sentence structure logic of planning constraints, the semantic relationships of land use classification terms, the numerical expression paradigm of development intensity constraints, the semantic orientation features of spatial boundary limitations, the intention expression features of planning demand texts, and the semantic differences of clauses at different constraint levels. Combining the requirements of demand analysis, rule extraction, and element identification, the training objectives of the model are set, and supervised tasks are constructed, including planning terminology recognition, planning intent classification, and land use demand element extraction. Each supervised task corresponds to the core function of the corresponding layer of the model. The planning terminology recognition task is used to train the model's planning knowledge fusion layer to recognize planning-specific professional terms (such as land use...). The model categorizes planning texts into several types, including: land use nature, plot ratio, building density, ecological protection boundaries, and urban development boundaries; planning intent classification; and land use demand element extraction. The latter trains the model's intent parsing output layer to classify various planning-related texts (e.g., land use potential assessment intent, development compliance assessment intent, supporting facility adaptation analysis intent, spatial layout rationality verification intent). This helps differentiate between different types of business demands. Finally, the model's intent parsing output layer extracts core elements required for assessment from text (e.g., assessment plot scope, target land use type, priority requirements, quantitative constraint indicators), enabling automated extraction of key demand information.

[0029] S203. Based on the supervised task, the initial model after incremental pre-training is fine-tuned under supervision. During the fine-tuning process, the importance of planning constraint level and land use assessment elements is combined to assign differentiated weights to different features in order to adjust the model parameters and optimize the feature extraction capability of the semantic coding layer, the correlation fusion capability of the planning knowledge fusion layer, and the result generation capability of the intent parsing output layer. Specifically, based on the initial model that has completed incremental pre-training, the initial model is fine-tuned using a constructed supervised task. The model's output is compared with the real labels labeled in the task, and the parameters of the three-layer architecture of the model are gradually optimized through error backpropagation. In addition, a dynamic feature weight allocation mechanism is introduced during the fine-tuning process. Based on the differences in the level of planning constraints (such as mandatory constraints and guiding constraints) and the importance of land use assessment elements (such as core assessment indicators and auxiliary reference elements), differentiated learning weights are assigned to different semantic features and text elements (such as assigning higher weights to mandatory constraint-related terms and lower weights to secondary auxiliary elements). Through differentiated weight adjustment, the model's planning knowledge fusion layer is strengthened to focus on the fusion of high-importance planning information, while the accuracy of the intent parsing output layer in extracting core elements and determining intent is improved. The interference of low-importance and low-relevance information is weakened, making the model's parsing results more in line with the hierarchical logic of planning constraints and the core requirements of land use assessment, thereby improving the professional adaptability and output accuracy of the three-layer architecture of the model.

[0030] S204. Simultaneously input the pre-built planning validation set and performance index evaluation rules into the supervised fine-tuning initial model for performance evaluation. If the model evaluation result does not reach the preset threshold, adjust the supervised fine-tuning parameters and repeat the fine-tuning process until the model performance meets the preset threshold to obtain the planning language processing model. Among them, the performance index evaluation rules include the completeness of planning term recognition, the accuracy of evaluation intent matching, and the recall rate of land use demand element extraction.

[0031] Specifically, a planning validation set (containing independent samples of various planning texts and demand texts) is pre-constructed that is from the same source as the planning training set but has no overlap with it. Simultaneously, performance evaluation rules are set, including the completeness of planning terminology recognition (measuring the model's planning knowledge fusion layer's ability to comprehensively recognize and integrate planning terminology), the accuracy of assessment intent matching (measuring the model's intent parsing output layer's ability to accurately classify land use assessment business intent), and the recall rate of land use demand element extraction (measuring the model's intent parsing output layer's ability to completely extract core assessment elements). The planning validation set and performance evaluation rules are then input synchronously. After supervised fine-tuning, the initial model undergoes an overall performance test, and the actual scores for each metric are calculated. If any performance metric fails to meet a preset threshold (e.g., term recognition completeness is less than 95%, intent matching accuracy is less than 93%), the hyperparameters of the supervised fine-tuning (e.g., learning rate, batch size, training epochs) are adjusted, and the supervised fine-tuning and performance evaluation process is repeated. Training and iteration are stopped until all performance metrics of the model meet the preset threshold requirements, resulting in a planning language processing model with a complete three-layer architecture of semantic encoding layer, planning knowledge fusion layer, and intent parsing output layer, which can be directly applied to requirement parsing and constraint extraction.

[0032] S30. Obtain multi-source planning land use data including spatial data and text data, preprocess the multi-source planning land use data to obtain a planning dataset.

[0033] Specifically, based on the demand-oriented parameters obtained in step S20 (such as assessment scope, core objectives, and constraint thresholds), two types of core multi-source planning land data are acquired: spatial data and textual data. Spatial data includes spatial attribute-related data for planned land, such as plot boundary coordinate vector data, remote sensing image data, topographic elevation data, urban development boundary spatial data, ecological protection boundary vector data, and land use functional zoning spatial data, used to reflect the core spatial characteristics of the plots, such as spatial location, shape, and spatial relationships. Textual data includes publicly available constraint rules and planning technical texts for planned land, specifically including… Planning constraint texts, planning spatial texts, and planning technical texts, such as the overall land space plan, regulatory detailed plan, land use zoning constraint requirements, land development constraint rules, development intensity control specifications, ecological protection constraint requirements, and public service facility configuration requirements for the corresponding region, are all from the same source and have highly overlapping content as the texts in the planning training set in step S20. They are all publicly available professional texts in the planning field. The only difference is that the planning training set is a generalization model training corpus, while the texts in this step are targeted measured data that match the land parcels to be evaluated, and are used to provide direct textual basis for the extraction of evaluation constraint parameters in S40.

[0034] The acquired multi-source planning land use data undergoes standardized preprocessing, including coordinate system unification, spatial correction, outlier removal, and overlay alignment of spatial data to ensure spatial data accuracy; invalid information removal, planning terminology normalization, and format unification of text data to improve the compatibility of text with planning language processing models; and a correlation mapping between spatial data and text data is established, such as binding the boundary coordinates of land parcels with the corresponding constraint text and planning technical text of the land parcels, to achieve collaborative matching of multi-source data and form a standardized planning dataset that can be directly used in subsequent evaluation processes.

[0035] It should be noted that all the aforementioned multi-source planning land use data are publicly available data that can be legally and compliantly obtained by those skilled in the art through public channels, and there are no non-public internal documents. Among them, spatial data can be queried and downloaded from public channels such as the official publicity platforms of natural resources authorities and geographic information public service platforms; the text data and the planning training set in step S20 are from the same source, and can be obtained from planning results publicized by natural resources authorities, government information disclosure websites, planning result filing systems, national and industry standard public service platforms, etc., which is clear to those skilled in the art.

[0036] S40. Guided by demand-oriented parameters, the planning dataset is parsed and extracted using a planning language processing model to generate evaluation constraint parameters associated with the demand-oriented parameters. The evaluation constraint parameters have rigid constraint rules and flexible guidance rules.

[0037] Specifically, guided by demand-oriented parameters, the system utilizes a planning language processing model to process unstructured data such as planning texts, planning constraint documents, and technical standards within the planning dataset. This process identifies and extracts key planning constraint information relevant to the assessment needs. Based on the constraint logic of the planning domain, the extracted information is hierarchically divided into rules, distinguishing between rigid constraint rules with mandatory force and flexible guiding rules with guiding force. The resulting rule information is then mapped to demand-oriented parameters, eliminating redundant rules irrelevant to the needs, supplementing missing constraints necessary for assessment, and generating assessment constraint parameters that contain dual rule attributes and are highly adapted to the needs.

[0038] S50. The demand-oriented parameters, evaluation constraint parameters and spatial data in the planning dataset are fused to obtain the evaluation input dataset.

[0039] Specifically, the spatial data in the demand-oriented parameters, evaluation constraint parameters, and planning datasets are standardized to unify the format and field naming conventions of various data types. Spatial association mapping between the three types of data is established based on the unique identifier information of each plot, realizing the collaborative binding of parameter information and spatial data. The bound data is then validated according to planning logic to identify fundamental conflicts between core evaluation objectives and rigid constraint rules and retain compliant bound data. Finally, the structured integration is completed according to the preset evaluation input template, redundant fields are removed, necessary attribute information is added, and a standardized evaluation input dataset suitable for subsequent model evaluation is formed.

[0040] S60. Input the assessment input dataset into the planning land potential assessment model, combine the preset assessment index system and weight allocation rules to conduct a comprehensive quantitative assessment, and output the planning land potential assessment result. The planning land potential assessment result includes at least a quantitative assessment score and a potential level label.

[0041] Specifically, the fused assessment input dataset is input into the pre-trained planning land potential assessment model. The model, based on the preset planning land potential assessment index system and combined with the weight allocation rules adapted to the needs, performs a hierarchical and quantitative comprehensive calculation of the compliance and potential value of the land parcels. The entire assessment process takes the rigid constraints of planning as the precondition for judgment, and then performs a weighted calculation of flexible potential, strictly following the constraint logic of territorial spatial planning, to ensure that the assessment results not only meet the constraint criteria, but also meet the core needs of users. Based on the comprehensive quantitative calculation results output by the planned land use potential assessment model, the results are standardized according to a preset format to generate quantifiable and comparable assessment scores and intuitively categorized potential level labels. The quantitative assessment score reflects the potential value level of the land parcel, while the potential level label is used to hierarchically classify the potential of the land parcel (such as high potential, medium potential, low potential, no potential, etc.). Finally, the planned land use potential assessment results, which contain the above core content, are output in a standardized format. The results have the characteristics of unified standards, traceability, and can be directly used for planning scheme preparation and land use decision-making, fully meeting the application needs of urban planning management and optimal allocation of land resources.

[0042] In one embodiment, step S40 specifically includes the following steps: S41. Extract the planning constraint text, planning spatial text and planning technical text from the planning dataset and preprocess them to obtain standardized planning text; S42. Input the standardized planning text and demand-oriented parameters into the planning language processing model. Based on the classification labels and core parameters in the demand-oriented parameters, extract the key assessment information from the standardized planning text. The key assessment information includes at least land use constraints, development intensity indicators, supporting facility requirements, and ecological constraint boundaries. S43. Based on the planning language processing model, hierarchical rule parsing is performed on the extracted key evaluation information to define the boundary between rigid constraint rules and flexible guidance rules, and generate rule parsing results. S44. Based on the constraint thresholds in the demand-oriented parameters, perform planning constraint conflict verification on the rule parsing results, eliminate contradictory constraint items, and complete missing constraint items to obtain the planning evaluation constraint information. S45. The comprehensive planning assessment constraint information is structured and mapped to the demand-oriented parameters to generate assessment constraint parameters that include constraint type labels, quantitative indicator thresholds, rule application scope, and priority ranking.

[0043] In this embodiment, step S41 specifically involves: extracting core texts related to planning constraints from the planning dataset, including planning constraint texts (such as control detailed planning constraint clauses, land use functional zoning constraint rules, etc.), planning spatial texts (such as area comprehensive planning texts, special planning constraint chapters, etc.), and planning technical texts (such as land development intensity control specifications, supporting facility configuration standards, etc.). Addressing issues such as inconsistent formats, semantic redundancy, and non-standard terminology in these three types of texts, multi-dimensional preprocessing operations are performed, including format unification (e.g., converting different formats such as PDF and Word into plain text format, unifying paragraph separators and encoding rules), redundancy removal (removing explanatory and descriptive content unrelated to constraint rules from the text), terminology normalization (unifying synonymous and heterogeneous constraint terms into standard expressions in the planning field, such as unifying building height limits and maximum building height into building height control), and missing information completion (completing missing constraint index values, applicable scope, and other key information in the text), resulting in standardized planning texts with unified format, standardized semantics, and complete information.

[0044] For step S42, specifically, the standardized planning text and the demand-oriented parameters generated in step S20 (such as the classification label "Residential Land Development Potential Assessment" and the core parameter "10 hectares of urban residential land in a certain district of a certain city") are simultaneously input into the planning language processing model. The model locks the assessment scenario based on the classification label in the demand-oriented parameters, clarifies the extraction scope through the core parameters, and focuses on extracting four types of key assessment information from the standardized planning text, including land use constraints (such as the land use of a certain area is urban residential land (R category), which cannot be changed to commercial land (B category)), development intensity indicators (such as plot ratio ≤ 2.5, building density ≤ 30%, building height ≤ 80 meters, etc.), supporting facility requirements (such as residential land must be equipped with a kindergarten, one per 10 hectares; community service center, building area ≥ 500 square meters), and ecological constraint boundaries (such as the ecological green belt within 50 meters west of the plot, where no building construction is allowed). During the extraction process, the model captures the correlation between the semantics of the text and the demand through the semantic encoding layer to ensure that the extracted information fully matches the user's assessment needs and avoids interference from irrelevant information.

[0045] For step S43, specifically, the model's planning knowledge fusion layer first associates the extracted key assessment information with the pre-set planning domain knowledge graph (e.g., associating ecological constraint boundaries with rigid constraints - ecological protection nodes in the knowledge graph), and performs rule hierarchy division based on planning constraint logic; then, the hierarchical rule parsing is executed through the intent parsing output layer; rules with mandatory binding force and cannot be broken are defined as rigid constraint rules (e.g., prohibiting any commercial construction activities within the ecological protection boundary, and prohibiting changes in land use across major categories, etc., violations will directly lead to unqualified assessment); rules with guiding nature and can be adjusted within a certain range are defined as flexible guiding rules (e.g., the plot ratio of residential land should be controlled between 1.8 and 2.5, and can be increased by 10% in special cases after demonstration; the location of supporting facilities can be appropriately adjusted according to the actual layout of the plot, etc.); during the parsing process, the applicable conditions and execution basis of the rules are recorded simultaneously (e.g., according to Article XX of the "Management Measures for Detailed Control Planning of a Certain City"), generating rule parsing results containing rule type, specific content, scope of application, and execution basis.

[0046] For step S44, specifically, based on the constraint thresholds in the demand-oriented parameters (such as plot ratio ≤ 2.5, ecological protection priority), the rule parsing results are subjected to multi-dimensional conflict verification. First, logical conflict verification (checking for contradictions between different rules, such as numerical conflicts between plot ratio ≤ 2.5 and plot ratio ≥ 3.0, and functional conflicts between allowing supporting commercial facilities and prohibiting non-residential supporting facilities). Conflicting rules are eliminated according to the principle of "rigidity over flexibility, and higher-level planning over lower-level planning" (e.g., retaining the rigid plot ratio ≤ 2.5 and eliminating the flexible plot ratio ≥ 3.0). Second, spatial conflict verification (checking for conflicts between the applicable scope of the rules and the scope of the assessed plot, such as removing rules applicable to "XX New Town" from the assessment constraints of the "XX Old Town" plot). Third, completeness verification (checking whether any core constraints necessary for the assessment are missing, such as supplementing key rules like requirements for supporting facilities and building height control when residential land assessments are missing). Through these three rounds of verification and processing, compliant assessment constraint information that is conflict-free, complete, and fully adaptable to the assessment needs is obtained.

[0047] For step S45, specifically, according to the parameterization requirements of the planning assessment business, the compliance assessment constraint information is structured and mapped; the rule type (rigid / flexible) is converted into constraint type labels (e.g., rigid - land use nature, flexible - development intensity); the specific values ​​of constraint indicators are converted into quantitative indicator thresholds (e.g., plot ratio ≤ 2.5, building height ≤ 80 meters, supporting kindergarten ≥ 1); the rule application scope is converted into rule application scope (e.g., a certain plot in a certain district of a certain city, plot number XXX); according to the planning constraint level and demand priority, the constraint items are prioritized (e.g., ecological constraint boundary priority is higher than supporting facility location adjustment, land use nature constraint priority is higher than plot ratio fluctuation); after field standardization and integration, the final assessment constraint parameters are generated, which include four major modules: constraint type labels, quantitative indicator thresholds, rule application scope, and priority ranking. These parameters are formatted uniformly and logically related with the demand-oriented parameters, and can be directly used for subsequent fusion processing with structured spatial data.

[0048] Through the above steps S41~S45, the pain points of traditional assessment, such as inefficiency in manual extraction of planning rules, ambiguity of rigid and elastic boundaries, and difficulty in identifying constraint conflicts, are resolved.

[0049] In one embodiment, step S50 specifically includes the following steps: S51. Perform field standardization processing on the spatial data in the demand-oriented parameters, evaluation constraint parameters and planning dataset to unify the data format and field naming rules; S52. Based on the unique identifier information of land parcels in spatial data, establish a spatial association mapping between demand-oriented parameters, evaluation constraint parameters and spatial data to obtain associated binding data; S53. Perform planning logic verification on the associated and bound data, verify whether there is a fundamental conflict between the core evaluation objectives in the demand-oriented parameters and the rigid constraint rules in the evaluation constraint parameters, and retain compliant bound data; S54. The compliance binding data is structured and integrated according to the preset assessment input field template, redundant fields are removed, necessary attribute information is added, and a unified assessment input dataset is formed.

[0050] In this embodiment, for step S51, specifically, to address the issues of inconsistent formats and chaotic field naming in the spatial data of the demand-oriented parameters, evaluation constraint parameters, and planning dataset, a standardization process is performed. First, the data format is unified: the text format of the demand-oriented parameters and evaluation constraint parameters is unified to JSON format, the vector format of the spatial data is unified to SHP format, and the coordinate system and data encoding rules are unified. Second, the field naming rules are unified: the plot number, parcel ID, and plot code are uniformly named as unique plot identifiers; the evaluation target and core requirements are uniformly named as core evaluation targets; and the constraint threshold and constraint index are uniformly named as quantitative index thresholds. This results in standardized multi-source data with unified format, standardized fields, and direct correlation.

[0051] For step S52, specifically, the unique identifier information of the land parcels in the spatial data of the planning dataset (such as parcel number XXX, coordinate code XXX, which is a unique identifier for each land parcel to be evaluated) is used as the association benchmark to establish a spatial association mapping relationship between the three types of data; the core evaluation objectives and evaluation scope in the demand-oriented parameters are bound to the unique identifier of the corresponding land parcel (such as binding the evaluation objective of residential land development potential to the spatial data of parcel number XXX), and the rigid constraint rules, flexible guidance rules, and quantitative indicator thresholds in the evaluation constraint parameters are bound to the unique identifier of the corresponding land parcel (such as binding rules such as plot ratio ≤ 2.5 and ecological green belt constraints to the spatial data of parcel number XXX); during the association process, spatial topology matching is used to assist in verification (such as verifying the association accuracy by the topological relationship between the land parcel range coordinates in the demand and the spatial data vector boundary) to avoid cross-parcel and wrong parcel association; through association mapping, a one-to-one correspondence between demand-oriented, constraint rules and land parcel spatial information is achieved, ensuring that each constraint parameter and each evaluation objective can be matched to the specific land parcel to be evaluated, and finally obtaining the association binding data of the three types of data collaboratively associated.

[0052] For step S53, specifically, a special planning logic verification is performed on the associated and bound data. The core verification content is whether there is a fundamental conflict between the core assessment objectives in the demand-oriented parameters and the rigid constraint rules in the assessment constraint parameters. Such conflicts are irreconcilable principle conflicts and cannot be resolved by adjusting flexible rules. For example, if the core assessment objective is "high-intensity development of commercial land" and the rigid constraint rule is "the land use nature of this plot is urban residential land and cannot be changed to commercial land", then there is a fundamental conflict between the two, and the associated and bound data of this plot needs to be removed. If the core assessment objective is "improvement of the supporting facilities of residential land" and the rigid constraint rule is "residential land must be equipped with kindergartens and community service centers", then there is no conflict between the two, and the associated and bound data is retained. During the verification process, the verification results and the reasons for the conflict are recorded simultaneously, and finally, compliant bound data that has no fundamental conflict and conforms to the planning constraint logic is selected.

[0053] For step S54, specifically, a preset evaluation input field template is invoked (the template includes core fields such as "unique land parcel identifier, core evaluation objective, constraint type label, quantitative indicator threshold, spatial coordinate information, and rule application scope") to structurally integrate the compliant binding data obtained in step S53. First, redundant fields are removed (such as fields irrelevant to the evaluation, such as data acquisition time and text format type). Second, necessary attribute information is supplemented (such as supplementing missing topographic elevation information in the land parcel spatial data and supplementing missing rule execution basis in the evaluation constraint parameters). Through field integration and standardization, it is ensured that all compliant binding data fields are unified and information is complete, forming a unified format evaluation input dataset that can be directly input into the planning land potential evaluation model.

[0054] Through the above steps S51~S54, the pain points of traditional data fusion, such as inconsistent formats, inaccurate data association, and fundamental constraint conflicts, are resolved, and efficient and compliant fusion of demand, constraints and spatial data is achieved.

[0055] In one embodiment, step S60, the construction and training of the planned land use potential assessment model, specifically includes the following steps: S601. Obtain historical land use assessment data, planning constraint rule execution data, and land use potential level labeling data in the planning field, and merge them with the pre-constructed planning training set to form a model training dataset. Perform standardized preprocessing and sample division on the model training dataset to obtain a training set, a validation set, and a test set. S602. Construct a model architecture that includes a feature adaptation layer, a planning rule fusion calculation layer and a potential level output layer. The feature adaptation layer is used to receive the evaluation input dataset, the planning rule fusion calculation layer integrates the evaluation index system and weight allocation rules, and the potential level output layer is used to output the quantitative evaluation results. S603. The model architecture is trained using the training set. During the training process, a hybrid loss function combining rigid planning constraints is adopted. The achievement of rigid compliance indicators is used as a prerequisite constraint. The weighted scoring results of the elastic potential indicators are optimized by backpropagation of errors, and the model parameters are iteratively adjusted. S604. Use the validation set and test set to evaluate the performance and generalization ability of the trained model. The evaluation indicators include the accuracy of potential level determination, the recall rate of rigid constraint verification, and the consistency of evaluation results. If the model performance does not meet the preset standard, adjust the model architecture parameters or training hyperparameters and repeat the training process until the model meets the preset performance requirements to obtain the planned land use potential evaluation model.

[0056] In this embodiment, for step S601, specifically, historical land use assessment data in the planning field (such as the potential assessment reports and quantitative score records of residential and commercial land in a certain city in the past 5 years), planning constraint rule execution data (such as statistics on the compliance of rigid constraint rules and case data of flexible rule adaptation), and land use potential level labeling data (such as high potential, medium potential, low potential, and no potential level labeling data labeled by experts, with labeling basis including compliance, development space, and supporting suitability, etc.); these three types of data are then integrated with the planning training set to supplement the label information and execution feedback data required for model training, forming a model training dataset.

[0057] Standardized preprocessing is performed on the model training dataset, including data format unification (e.g., converting text-based rule data into structured features and encoding level labels into numerical labels, such as high potential = 4, medium potential = 3, low potential = 2, no potential = 1), outlier removal (e.g., removing invalid samples with contradictory evaluation conclusions or missing data rates exceeding 30%), and missing value completion (e.g., using planning techniques to complete missing constraint thresholds). Samples are divided according to a preset ratio, such as training set: validation set: test set = 7:2:1, to ensure consistent sample distribution across datasets, avoid model overfitting, and obtain training, validation, and test sets that can be directly used for model training.

[0058] For step S602, specifically, the basic architecture of the model is constructed, including the feature adaptation layer, the planning rule fusion calculation layer, and the potential level output layer; The feature adaptation layer is used to receive the evaluation input dataset and normalize, map dimensions, and enhance features such as basic identifiers, requirement information, constraint information, and spatial attributes in the evaluation input dataset (e.g., converting spatial coordinate features into grid-coded features and text-type constraint features into vector features) to ensure that the input features are adapted to the model calculation. The planning rule integration calculation layer is the core layer of the model. It integrates a pre-built evaluation index system and weight allocation rules. It achieves the integration of index, weight and rule through preset calculation logic (such as first performing item-by-item verification and weighted summation of rigid compliance indicators, and then performing weighted scoring of flexible potential indicators based on compliance results). At the same time, it embeds planning domain-specific logic (such as the judgment logic of veto for rigid indicators and the gradient scoring logic of flexible indicators). The potential level output layer adopts a structure that combines a multi-classifier and a regressor to output two types of core quantitative evaluation results: one is the potential level label (such as high potential, medium potential, low potential, no potential), and the other is the specific quantitative score (such as a potential score on a scale of 0-100, where the higher the score, the greater the potential), in order to meet the output needs of different evaluation scenarios.

[0059] For step S603, specifically, the training set divided according to a preset ratio is input into the constructed model infrastructure to train the model; wherein, during the training process, a hybrid loss function combining planning rigid constraints is used, and this function consists of two parts: The first is the rigid constraint loss term (used to penalize the model for misjudging rigid compliance indicators, such as the error of classifying unqualified land parcels as compliant). Second is the elasticity potential loss term (used to optimize the quantitative scoring accuracy of the elasticity potential index, such as the deviation between the predicted score and the actual score).

[0060] Training is pre-constrained by meeting rigid compliance indicators. The model must first pass a rigid constraint verification logic before weighting and scoring the elastic potential indicators of compliant samples. If a sample fails to meet the rigid indicators, it is directly judged as having no potential and does not need to proceed to the potential scoring stage. Using an error backpropagation algorithm, model parameters (such as the weight matrix of the feature adaptation layer and the logical coefficients of the planning rule fusion calculation layer) are iteratively adjusted based on the calculation results of the hybrid loss function. Training hyperparameters (such as learning rate = 0.001, batch size = 32, and iteration times = 100) are set until the loss function value on the training set converges to a preset range.

[0061] For step S604, specifically, the performance of the trained model is evaluated using the validation set, focusing on three core evaluation metrics: First, the accuracy of potential level determination, which measures the degree of match between the potential level predicted by the model and the actual level, such as a preset standard of ≥92%; Second, rigid constraint verification recall rate is used to measure the model's ability to identify rigid violation samples. For example, the preset standard is ≥95%, which ensures that no rigid violation samples are misjudged as compliant. Third, the consistency of evaluation results is used to measure the stability of the model's evaluation results for similar samples. For example, the coefficient of variation of the evaluation scores of land parcels of the same type and under the same constraints should be ≤5%. If the validation set evaluation results do not meet the preset standards, adjust the model architecture parameters, such as increasing the number of network layers in the feature adaptation layer, optimizing the logical weights of the planning rule fusion calculation layer; or adjust the training hyperparameters, such as adjusting the learning rate to 0.0005 and increasing the number of iterations to 150, and repeat the training process. Once the validation set meets the standards, the model's generalization ability is validated using the test set to test its adaptability to new, unseen samples. If the test set indicators also meet the preset standards, the iteration stops, thus obtaining a planning land use potential assessment model that conforms to the planning constraint logic.

[0062] In one embodiment, step S60, the comprehensive quantitative evaluation specifically includes the following steps: S61. Input the assessment input dataset into the planning land use potential assessment model, and complete the multi-source data format adaptation and feature extraction through the feature adaptation layer to generate standardized model input features; S62. The planning rule fusion calculation layer performs rigid compliance indicator verification on the input features of the standardized model based on the evaluation indicator system and weight allocation rules. If any rigid compliance indicator fails to meet the standard, the land is directly judged as having no potential. If all rigid compliance indicators meet the standard, the flexible potential indicators are weighted and scored based on the weight allocation rules to obtain a quantitative score of flexible potential. S63. The potential level output layer generates a planned land use potential level label for the plot based on the elastic potential quantitative score and the preset potential level classification threshold, and outputs the quantitative assessment score and potential level label to complete the comprehensive quantitative assessment.

[0063] In this embodiment, for step S61, specifically, the evaluation input dataset containing fields such as basic identifier, demand information, constraint information, and spatial attributes is input into the planning land potential evaluation model; the model's feature adaptation layer first performs data format adaptation, converting different types of input data, such as text-type constraint labels, numerical index thresholds, and vector-type spatial boundaries, into a model-compatible tensor format, such as encoding rigid-land use nature labels as the number 1 and flexible-development intensity as the number 2.

[0064] Then, multi-dimensional feature extraction is performed to extract the core features corresponding to the evaluation index system from the input data, such as extracting the land area, shape coefficient, and coordinate boundary features from the spatial attribute field, and extracting rigid / elastic identifiers and quantization threshold features from the constraint information field.

[0065] Finally, feature standardization is performed. Through normalization and dimension alignment, the impact of differences in feature dimensions on the evaluation results is eliminated. For example, continuous features such as area and distance are mapped to the [0,1] interval to ensure that the feature dimension is consistent with the model's preset input dimension. Standardized model input features with unified structure, adapted dimensions, and can be directly used for model calculation are generated.

[0066] For step S62, specifically, the model's planning rule fusion calculation layer performs two core calculations based on the evaluation index system and weight allocation rules: First, there is a rigid compliance indicator verification process. The three dimensions of land use constraint compliance, ecological constraint matching degree, and planning boundary compliance are verified and scored one by one according to the basic weight. For example, 40 points are awarded for meeting the land use constraint compliance standard and 0 points are awarded for not meeting it. 35 points are awarded for meeting the ecological constraint matching degree and 0 points are awarded for not meeting it. The total rigid compliance score is obtained by summing the scores. For example, 40+35+25=100 points is considered fully compliant. If any rigid dimension fails to meet the standard, such as the ecological constraint matching degree receiving 0 points, a veto mechanism is triggered directly, and the land is judged as a land parcel with no potential, and the subsequent calculation is terminated.

[0067] Second, a weighted scoring system based on flexibility potential indicators is used only for plots that meet rigid compliance standards, such as those with a total rigid score ≥ 100 points, or those judged as compliant according to a preset compliance threshold. Based on the dynamic weights corresponding to demand-oriented parameters, scores are awarded separately for three dimensions: development intensity adaptability, supporting facility completeness, and spatial layout rationality. For example, 45 points are awarded for development intensity suitability, 30 points for the completeness of supporting facilities, and 25 points for the rationality of spatial layout. The weighted summation yields the quantitative score of elastic potential, such as 45×45%+30×30%+25×25%=35.5 points. The specific calculation logic is executed according to the model's preset. The final output is the rigid compliance judgment result and the quantitative score of elastic potential. Among them, only the judgment result is output for plots with no potential.

[0068] For step S63, specifically, the potential level output layer of the model is pre-set with potential level classification thresholds, and calibrated based on the experience of planning experts and historical evaluation data. For example, under the 0-100 score system: 0 points = no potential, 1-30 points = low potential, 31-60 points = medium potential, 61-100 points = high potential. The obtained elasticity potential score, such as 35.5 points, is compared with a preset threshold to generate a corresponding potential level label, such as 35.5 points corresponding to medium potential; at the same time, two types of core evaluation results are integrated and output: The first is the quantitative assessment score, which includes the total score for rigid compliance, the quantitative score for flexibility potential, and the comprehensive score. For example, the rigid score is 100 points, the flexibility score is 35.5 points, and the comprehensive score = rigid score × rigid-flexibility ratio weight + flexibility score × rigid-flexibility ratio weight, that is, 100 × 70% + 35.5 × 30% = 80.65 points. Second, potential level labels, such as medium potential; the output results can be presented in preset formats, such as structured reports, JSON data, and visual charts, clearly marking the evaluation basis, such as the specific dimensions of rigid compliance and the core contribution dimensions of flexible scoring, so as to complete the comprehensive quantitative evaluation of the planning land potential of the plot.

[0069] In one embodiment, step S60, the construction of the preset evaluation index system, specifically includes the following steps: S605. Based on the core objective of the planning land use potential assessment, the assessment indicator system is divided into rigid compliance indicators and flexible potential indicators, and the assessment priority and judgment logic of rigid compliance indicators and flexible potential indicators are set. S606. The specific evaluation dimensions for configuring rigid compliance indicators shall include at least land use constraint compliance, ecological constraint matching degree and planning boundary compliance, and each evaluation dimension shall correspond to the rigid constraint rules in the evaluation constraint parameters. S607. The specific evaluation dimensions of the configuration flexibility potential indicators shall include at least the adaptability of development intensity, the completeness of supporting facilities and the rationality of spatial layout, and each evaluation dimension shall be related to the core evaluation objectives in the demand-oriented parameters.

[0070] In this embodiment, for step S605, specifically, the evaluation index system is first divided into two core categories: Rigid compliance indicators and flexible potential indicators; rigid compliance indicators are used to determine whether a plot of land meets the planning constraints that cannot be broken, and are a prerequisite for passing the assessment; flexible potential indicators are used to quantify the development potential and adaptability of a plot of land on the basis of compliance. Furthermore, a clear assessment priority is set: rigid compliance indicators > flexible potential indicators; that is, rigid compliance verification must be passed first before flexible potential is quantified. If rigid indicators are not met, the assessment is directly deemed unqualified and there is no need to proceed to the potential assessment stage. Finally, the judgment logic is set as follows: rigid compliance indicators adopt a veto system + full compliance system. If any rigid dimension is not met, the overall compliance is judged as unqualified. Flexible potential indicators adopt a multi-dimensional weighted scoring system. The higher the score, the greater the potential, the more flexible dimensions are calculated into a comprehensive score.

[0071] For step S606, specifically, three core evaluation dimensions are configured, and each dimension is directly associated with the rigid constraint rules in the evaluation constraint parameters: First, land use constraints and compliance, focusing on core rigid requirements such as land use nature and development intensity. For example, verifying whether the actual land use nature of the plot is consistent with the rigid rule that urban residential land (R category) cannot be changed, whether the plot ratio meets the rigid threshold of ≤2.5, and whether the building density meets the mandatory requirement of ≤30%. Second, ecological constraint matching degree, benchmarking against the rigid constraint rules related to ecological protection, such as verifying whether the plot is located in the rigid area where construction is prohibited within the ecological protection boundary, whether the green space ratio reaches the mandatory standard of ≥30%, and whether it meets the constraint requirement that there are no buildings in the 50-meter ecological green belt on the west side of the plot; Third, the compliance with planning boundaries is checked. The degree of conformity between the plot and the statutory planning boundary is verified, such as whether the plot is within the urban development boundary, whether it meets the rigid requirements of the functional zoning of the area, and whether it violates the mandatory provision that the plot boundary shall not cross the municipal road boundary. Each assessment dimension has a clear compliance judgment standard, such as 100 points for compliance and 0 points for non-compliance, to ensure that the rigid verification results are clear and quantifiable.

[0072] For step S607, specifically, three core evaluation dimensions are configured, and each dimension is strongly correlated with the core evaluation objectives in the demand-oriented parameters: First, development intensity adaptability focuses on the optimization space of development intensity. For example, in response to the need for assessing the development potential of residential land, the adaptability of the current plot ratio of 1.8 to the flexible guidance rule of 1.8-2.5 is verified, and the potential score corresponding to the space that can be improved is calculated. The higher the adaptability, the higher the score. Second, the completeness of supporting facilities, which is related to users' potential demand for supporting services. For example, for the assessment of residential land, the coverage radius of supporting facilities such as kindergartens, primary schools, and community service centers around the plot is checked. For example, if the service radius of kindergartens is ≤300 meters and the construction scale meets the flexible guidance rules, the completeness is scored as follows: full coverage = 100 points, partial coverage = 60 points, and no coverage = 20 points. Third, the rationality of spatial layout, assess the compatibility of land use with the surrounding environment, such as assessing the development potential of commercial land, verifying whether the shape of the land is suitable for the layout of commercial buildings, whether the distance to surrounding transportation hubs, such as subway entrances, meets the flexible guidance requirement of ≤500 meters, and whether the functional complementarity with surrounding residential areas meets the standard. Each evaluation dimension has a tiered scoring standard to ensure that the potential quantification results are aligned with the user's core needs and the flexible guidance rules for planning.

[0073] In one embodiment, step S60, the construction of the preset weight allocation rule, specifically includes the following steps: S608. Based on the assessment dimensions of planning constraint levels and rigid compliance indicators, assign basic weights to each assessment dimension of rigid compliance indicators; wherein, the sum of the basic weights of each assessment dimension is a fixed value, and the basic weights of land use constraint compliance and ecological constraint matching degree are not lower than the basic weights of planning boundary compliance. S609. Based on the priority of the core assessment objectives in the demand-oriented parameters, assign dynamic adjustment weights to each assessment dimension of the elastic potential indicator; wherein, the sum of the dynamic adjustment weights of each assessment dimension is a fixed value, the weight of the elastic potential assessment dimension corresponding to the core assessment objective is higher than that of the non-core assessment dimension, and the sum of the dynamic adjustment weights of each assessment dimension of the elastic potential indicator forms a preset proportional relationship with the sum of the basic weights of each assessment dimension of the rigid compliance indicator.

[0074] In this embodiment, for step S608, specifically, the three configured rigid compliance assessment dimensions are used as the allocation objects, and the weight allocation is based on two core principles: First, the principle of correlation between planning constraint levels applies: the higher the constraint level of a dimension (corresponding to the more rigid constraint rules in the assessment constraint parameters), the higher its basic weight. For example, the ecological protection boundary constraint belongs to the legally binding rules at the provincial level or above, and the corresponding ecological constraint matching degree dimension has a higher weight than the general constraint dimension. Second, the principle of statutory priority is strictly followed, ensuring that the basic weights of land use constraint compliance and ecological constraint matching are no less than those of planning boundary compliance, highlighting the importance of core rigid bottom lines such as land use nature and ecological protection; the sum of the basic weights of each assessment dimension is set as a fixed value, such as 100 under a 100-point system or 1.0 under a proportional system, and specific weights are allocated in accordance with the above principles; for example, under the proportional system: land use constraint compliance 40%, ecological constraint matching 35%, planning boundary compliance 25%; under the 100-point system: land use constraint compliance 40 points, ecological constraint matching 35 points, planning boundary compliance 25 points; this basic weight is a fixed value and will not be adjusted with changes in assessment needs, ensuring that the bottom line standards are consistent across all rigid compliance verification scenarios.

[0075] For step S609, specifically, the three configured elasticity potential assessment dimensions are used as the allocation objects, and the weight allocation follows three core rules: First, there is the demand priority matching rule. The elasticity dimension directly corresponding to the core assessment objective has a higher weight than the non-core dimension. For example, if the core assessment objective in the demand-oriented parameters is to maximize the development potential of residential land, then the development intensity adaptability dimension has the highest weight; if the core objective is to optimize the supporting services for commercial land, then the supporting facilities completeness dimension has the highest weight. Second, there is a fixed total weighting rule. The sum of the dynamically adjusted weights for each flexible dimension is a fixed value, such as 100 for a 100-point scale or 1.0 for a proportional scale. For example, when the core objective is to assess the development potential of residential land, the weight allocation is 45% for development intensity suitability, 30% for the completeness of supporting facilities, and 25% for the rationality of spatial layout. When the core objective is to optimize the spatial layout of commercial land, the weight allocation is 45% for the rationality of spatial layout, 35% for the completeness of supporting facilities, and 20% for the suitability of development intensity. Third, there are pre-set rules for the ratio of rigidity and flexibility. The sum of the weights of each dimension of the flexible potential indicator and the sum of the weights of each dimension of the rigid compliance indicator form a pre-set fixed ratio relationship. For example, the sum of rigid weights: the sum of flexible weights = 7:3 or 6:4. The specific ratio is preset according to the business scenario of planning and evaluation. For example, 8:2 is used in ecologically sensitive areas with extremely high compliance requirements, and 6:4 is used in urban development boundaries with prominent potential mining needs. By dynamically allocating weights, we can accurately emphasize potential dimensions under different assessment needs. At the same time, by controlling the ratio of rigidity to flexibility, we can ensure that the compliance bottom line is not weakened and the potential assessment does not deviate.

[0076] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0077] In one embodiment, an urban planning land use potential assessment system is provided, which corresponds to the urban planning land use potential assessment method described in the above embodiments. This urban planning land use potential assessment system includes: The demand response module is used to respond to the natural language demands input by the user, preprocess the natural language demands, and obtain the planning demand text. The requirement extraction module is used to input the planning requirement text into the planning language processing model, identify the user's planning evaluation intention type and extract the core evaluation objectives and additional constraints, integrate the planning evaluation intention type, core evaluation objectives and additional constraints, and generate requirement-oriented parameters containing classification labels, core parameters and constraint thresholds. The data acquisition module is used to acquire multi-source planned land use data, including spatial data and text data, and to preprocess the multi-source planned land use data to obtain a planning dataset. The constraint extraction module is used to parse and extract planning datasets using a planning language processing model, guided by demand-oriented parameters, to generate evaluation constraint parameters associated with the demand-oriented parameters. The evaluation constraint parameters have rigid constraint rules and flexible guidance rules. The data fusion module is used to fuse the demand-oriented parameters, evaluation constraint parameters and spatial data in the planning dataset to obtain the evaluation input dataset; The evaluation output module is used to input the evaluation input dataset into the planning land potential evaluation model, combine it with the preset evaluation index system and weight allocation rules to conduct a comprehensive quantitative evaluation, and output the planning land potential evaluation result. The planning land potential evaluation result includes at least a quantitative evaluation score and a potential level label.

[0078] For specific limitations regarding an urban planning land use potential assessment system, please refer to the limitations of an urban planning land use potential assessment method described above, which will not be repeated here. Each module in the aforementioned urban planning land use potential assessment system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0079] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows. Figure 2As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database is used for data storage, data processing, and data analysis. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for assessing the potential of urban planning land use.

[0080] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a method for assessing the potential of urban planning land use.

[0081] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements a method for assessing the potential of urban planning land use.

[0082] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0083] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0084] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for assessing the potential of urban planning land use, characterized in that, Includes the following steps: S10. In response to the user's input of natural language requirements, the natural language requirements are preprocessed to obtain the planning requirement text; S20. Input the planning requirement text into the planning language processing model, identify the user's planning evaluation intention type and extract the core evaluation objectives and additional constraints, integrate the planning evaluation intention type, core evaluation objectives and additional constraints to generate a demand-oriented parameter that includes classification labels, core parameters and constraint thresholds. S30. Obtain multi-source planned land use data including spatial data and text data, and preprocess the multi-source planned land use data to obtain a planning dataset; S40. Guided by demand-oriented parameters, the planning dataset is parsed and extracted using a planning language processing model to generate evaluation constraint parameters associated with the demand-oriented parameters. The evaluation constraint parameters have rigid constraint rules and flexible guidance rules. S50. The demand-oriented parameters, evaluation constraint parameters and spatial data in the planning dataset are fused to obtain the evaluation input dataset; S60. Input the assessment input dataset into the planning land potential assessment model, combine the preset assessment index system and weight allocation rules to conduct a comprehensive quantitative assessment, and output the planning land potential assessment result. The planning land potential assessment result includes at least a quantitative assessment score and a potential level label.

2. The method for assessing urban planning land use potential as described in claim 1, characterized in that, Step S40 specifically includes the following steps: S41. Extract the planning constraint text, planning spatial text and planning technical text from the planning dataset and preprocess them to obtain standardized planning text; S42. Input the standardized planning text and demand-oriented parameters into the planning language processing model. Based on the classification labels and core parameters in the demand-oriented parameters, extract the key assessment information in the standardized planning text. The key assessment information includes at least land use constraints, development intensity indicators, supporting facility requirements, and ecological constraint boundaries. S43. Based on the planning language processing model, hierarchical rule parsing is performed on the extracted key evaluation information to define the boundary between rigid constraint rules and flexible guidance rules, and generate rule parsing results. S44. Based on the constraint thresholds in the demand-oriented parameters, perform planning constraint conflict verification on the rule parsing results, eliminate contradictory constraint items, and complete missing constraint items to obtain the planning evaluation constraint information. S45. The planning assessment constraint information and demand-oriented parameters are structured and mapped to generate assessment constraint parameters that include constraint type labels, quantitative indicator thresholds, rule application scope and priority ranking.

3. The method for assessing urban planning land use potential as described in claim 1, characterized in that, Step S50 specifically includes the following steps: S51. Perform field standardization processing on the spatial data in the demand-oriented parameters, evaluation constraint parameters and planning dataset to unify the data format and field naming rules; S52. Based on the unique identifier information of land parcels in spatial data, establish a spatial association mapping between demand-oriented parameters, evaluation constraint parameters and spatial data to obtain associated binding data; S53. Perform planning logic verification on the associated and bound data, verify whether there is a fundamental conflict between the core evaluation objectives in the demand-oriented parameters and the rigid constraint rules in the evaluation constraint parameters, and retain compliant bound data; S54. The compliance binding data is structured and integrated according to the preset assessment input field template, redundant fields are removed, necessary attribute information is added, and a unified assessment input dataset is formed.

4. The method for assessing urban planning land use potential as described in claim 1, characterized in that, Step S60 also includes the construction and training of the planned land use potential assessment model, specifically including the following steps: S601. Obtain historical land use assessment data, planning constraint rule execution data, and land use potential level labeling data in the planning field, and merge them with the pre-constructed planning training set to form a model training dataset. Perform standardized preprocessing and sample division on the model training dataset to obtain a training set, a validation set, and a test set. S602. Construct a model architecture that includes a feature adaptation layer, a planning rule fusion calculation layer and a potential level output layer. The feature adaptation layer is used to receive the evaluation input dataset, the planning rule fusion calculation layer integrates the evaluation index system and weight allocation rules, and the potential level output layer is used to output the quantitative evaluation results. S603. The model architecture is trained using the training set. During the training process, a hybrid loss function combining rigid planning constraints is adopted. The achievement of rigid compliance indicators is used as a prerequisite constraint. The weighted scoring results of the elastic potential indicators are optimized by backpropagation of errors, and the model parameters are iteratively adjusted. S604. Use the validation set and test set to evaluate the performance and generalization ability of the trained model. The evaluation indicators include the accuracy of potential level determination, the recall rate of rigid constraint verification, and the consistency of evaluation results. If the model performance does not meet the preset standard, adjust the model architecture parameters or training hyperparameters and repeat the training process until the model meets the preset performance requirements to obtain the planned land use potential evaluation model.

5. The method for assessing urban planning land use potential as described in claim 4, characterized in that, In step S60, the comprehensive quantitative evaluation specifically includes the following steps: S61. Input the assessment input dataset into the planning land use potential assessment model, and complete the multi-source data format adaptation and feature extraction through the feature adaptation layer to generate standardized model input features; S62. The planning rule fusion calculation layer performs rigid compliance indicator verification on the input features of the standardized model based on the evaluation indicator system and weight allocation rules. If any rigid compliance indicator fails to meet the standard, the land is directly judged as having no potential. If all rigid compliance indicators meet the standard, the flexible potential indicators are weighted and scored based on the weight allocation rules to obtain a quantitative score of flexible potential. S63. The potential level output layer generates a planned land use potential level label for the plot based on the elastic potential quantitative score and the preset potential level classification threshold, and outputs the quantitative assessment score and potential level label to complete the comprehensive quantitative assessment.

6. The method for assessing urban planning land use potential as described in claim 1, characterized in that, In step S60, the construction of the preset evaluation index system specifically includes the following steps: S605. Based on the core objective of the planning land use potential assessment, the assessment indicator system is divided into rigid compliance indicators and flexible potential indicators, and the assessment priority and judgment logic of rigid compliance indicators and flexible potential indicators are set. S606. The specific evaluation dimensions for configuring rigid compliance indicators shall include at least land use constraint compliance, ecological constraint matching degree and planning boundary compliance, and each evaluation dimension shall correspond to the rigid constraint rules in the evaluation constraint parameters. S607. The specific evaluation dimensions of the configuration flexibility potential indicators shall include at least the adaptability of development intensity, the completeness of supporting facilities and the rationality of spatial layout, and each evaluation dimension shall be related to the core evaluation objectives in the demand-oriented parameters.

7. The method for assessing urban planning land use potential as described in claim 6, characterized in that, In step S60, the construction of the weight allocation rule specifically includes the following steps: S608. Based on the assessment dimensions of planning constraint levels and rigid compliance indicators, assign basic weights to each assessment dimension of rigid compliance indicators; wherein, the sum of the basic weights of each assessment dimension is a fixed value, and the basic weights of land use constraint compliance and ecological constraint matching degree are not lower than the basic weights of planning boundary compliance. S609. Based on the priority of the core assessment objectives in the demand-oriented parameters, assign dynamic adjustment weights to each assessment dimension of the elastic potential indicator; wherein, the sum of the dynamic adjustment weights of each assessment dimension is a fixed value, the weight of the elastic potential assessment dimension corresponding to the core assessment objective is higher than that of the non-core assessment dimension, and the sum of the dynamic adjustment weights of each assessment dimension of the elastic potential indicator forms a preset proportional relationship with the sum of the basic weights of each assessment dimension of the rigid compliance indicator.

8. A system for assessing the potential of urban planning land, used to implement the steps of the method for assessing the potential of urban planning land as described in any one of claims 1-7, characterized in that, include: The demand response module is used to respond to the natural language demands input by the user, preprocess the natural language demands, and obtain the planning demand text. The requirement extraction module is used to input the planning requirement text into the planning language processing model, identify the user's planning evaluation intention type and extract the core evaluation objectives and additional constraints, integrate the planning evaluation intention type, core evaluation objectives and additional constraints, and generate requirement-oriented parameters containing classification labels, core parameters and constraint thresholds. The data acquisition module is used to acquire multi-source planned land use data, including spatial data and text data, and to preprocess the multi-source planned land use data to obtain a planning dataset. The constraint extraction module is used to parse and extract planning datasets using a planning language processing model, guided by demand-oriented parameters, to generate evaluation constraint parameters associated with the demand-oriented parameters. The evaluation constraint parameters have rigid constraint rules and flexible guidance rules. The data fusion module is used to fuse the demand-oriented parameters, evaluation constraint parameters and spatial data in the planning dataset to obtain the evaluation input dataset; The evaluation output module is used to input the evaluation input dataset into the planning land potential evaluation model, combine it with the preset evaluation index system and weight allocation rules to conduct a comprehensive quantitative evaluation, and output the planning land potential evaluation result. The planning land potential evaluation result includes at least a quantitative evaluation score and a potential level label.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the urban planning land use potential assessment method as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for assessing urban planning land use potential as described in any one of claims 1-7.

Images