Land space planning evaluation method based on deep learning

Abstract

The invention belongs to the technical field of space planning, and particularly relates to a land space planning evaluation method based on deep learning, and the method comprises the steps: constructing a scene graph structure, generating an initial scene graph, updating the scene graph through an iteration planner, recording and feeding back, and enabling the scene graph structure to visually present the relation between planning objects. The complex territorial space layout can be conveniently analyzed; the method comprises the steps of extracting physical characteristics of a current scene graph, deploying a planning evaluation model to generate label distribution, determining evaluation indexes to calculate spatial planning loss, optimizing the planning evaluation model, deeply mining current scene graph information, facilitating accurate grasping of relation characteristics between objects, providing accurate evaluation for spatial planning, comprehensively considering planning constraints, and improving the planning efficiency. Reliable decision support is provided; a dynamic updating mechanism is adopted to discover a space planning problem in time, and rationality and feasibility of space planning are improved.

Description

Technical Field

[0001] The present invention belongs to the field of spatial planning technology, and specifically refers to a national land space planning evaluation method based on deep learning. Background Art

[0002] National land space planning refers to the spatial and temporal arrangements for the development and protection of national land space in a certain area. It is the basic basis for various development, protection and construction activities. The current national land space planning has the following shortcomings:

[0003] The current national land space planning process is difficult to grasp the complex spatial relationship. In the regional development plan, it is impossible to clearly show the mutual connection and influence between different functional areas, and it is difficult to adjust the plan in time according to the actual development situation, which may cause confusion in the layout of functional areas and uncoordinated infrastructure construction. For example, there is a lack of reasonable transportation connection and supporting facilities planning between industrial areas and residential areas;

[0004] The current national land space planning is unable to deeply analyze the spatial layout characteristics, and thus cannot accurately evaluate the planning effect, resulting in the discovery of low land use efficiency and ecological environmental damage after the planning scheme is implemented. In the planning process, the coordinated development of economy, society and ecology cannot be fully considered, and the excessive pursuit of economic growth and neglect of ecological environmental protection have led to over-exploitation of land resources.

[0005] The current national land space planning lacks a long-term strategic layout and only meets current short-term needs, which is not conducive to sustainable development. Especially in transportation planning, future urban expansion and changes in transportation patterns are not taken into account, resulting in planning strategies that cannot adapt to development needs and require frequent transformation and expansion, resulting in waste of resources. In addition, planning strategies lack flexibility and agility and are unable to cope with uncertainties and changes. Summary of the invention

[0006] In view of the above problems, the present invention provides a land space planning evaluation method based on deep learning. The method constructs a scene graph structure, generates an initial scene graph, uses an iterative planner to update the scene graph and record feedback. The scene graph structure can intuitively present the relationship between planning objects, which is convenient for analyzing complex land space layout; extracts the physical characteristics of the current scene graph, deploys a planning evaluation model to generate label distribution, determines evaluation indicators to calculate spatial planning losses, optimizes the planning evaluation model, and deeply mines the current scene graph information, which helps to accurately grasp the relationship characteristics between objects, provide accurate evaluation for spatial planning, comprehensively consider planning constraints, improve model robustness and generalization ability, and provide reliable decision support; adopts a dynamic update mechanism to timely discover spatial planning problems and improve the rationality and feasibility of spatial planning.

[0007] A land space planning evaluation method based on deep learning specifically includes the following steps:

[0008] Step S1: data collection and preprocessing: collecting planning data of national land space planning, performing preprocessing operations on the planning data of national land space planning, obtaining preprocessed data, and generating a global dictionary;

[0009] Step S2: scene graph construction and update: construct the initial scene graph, generate the planned action sequence and planning constraints, execute the planned action sequence in compliance with the planning constraints, convert the initial scene graph into the current scene graph, and generate feedback information of the planned action sequence;

[0010] Step S3: Planning evaluation model deployment: extract the physical features of the current scene graph, build a planning evaluation model, obtain the label distribution of the current scene graph according to the physical features, determine the planning evaluation index, and further calculate the spatial planning loss. By minimizing the spatial planning loss, train and optimize the planning evaluation model;

[0011] Step S4: Spatial planning assessment application: Use the planning assessment model to conduct land use assessment, traffic condition assessment and socio-economic assessment, and generate an assessment result report.

[0012] Furthermore, step S1 includes the following steps:

[0013] Step S11: Collecting geospatial data: Collecting high-resolution satellite images, and using satellite images to provide land use type information and topographic information. The land use type information includes construction land, agricultural land and ecological land, and the topographic information includes altitude, slope and slope direction;

[0014] Step S12: Collecting social and economic data: Collecting population distribution data, economic development indicators and transportation network distribution information, where the economic development indicators include GDP and industrial structure data, and the transportation network distribution information includes flow data;

[0015] Step S13: Collect planning document data: collect planning objectives and spatial layout plans of national land space planning. Planning objectives include the proportion of ecological protection red lines and construction land scale control targets. Spatial layout plans include functional zoning and industrial layout.

[0016] Step S14: pre-processing geospatial data: using ground control points, correcting the image to the correct geographic coordinate system through a polynomial fitting method, normalizing the geospatial data, and mapping the value range to a specific interval;

[0017] Step S15: preprocessing socioeconomic data: using K-nearest neighbor algorithm to fill missing values ​​in socioeconomic data;

[0018] Step S16: Preprocessing planning file data: extracting key indicators and time nodes in planning objectives through natural language processing technology, and vectorizing the spatial layout plan;

[0019] Step S17: Construct a global dictionary: merge the names extracted from the planning data of the national land space planning to form an initial name set, and delete duplicate names to generate a global dictionary.

[0020] Further, step S15 includes the following steps:

[0021] Step S151: Processing missing values: For the missing flow data of certain road sections in the traffic network distribution information, the K-nearest neighbor algorithm is used to predict the missing flow data values ​​based on the flow data and flow data of surrounding road sections;

[0022] Step S152: Normalization processing: converting the data into a standard normal distribution form;

[0023] Step S153: Multi-source data fusion: Population distribution data, economic development indicators and transportation network distribution information are fused through weighted summation and feature splicing.

[0024] Further, step S2 includes the following steps:

[0025] Step S21: define the scene graph: construct a planning scene graph structure G={V,E} based on the preprocessed data, where V is a node representing an object in the planning scene, and E is an edge describing the relationship between objects. The relationship set is constructed as R, representing the basic spatial planning relationship;

[0026] Step S22: Generate a scene graph: Use a scene graph generator to receive preprocessed data, a relationship set and a global dictionary according to the planned scene graph structure, and generate an initial scene graph, wherein the initial scene graph includes N objects;

[0027] Step S23: Define the planning task: According to the initial scene graph, design a spatial planner based on Transformer technology to generate a planning action sequence and planning constraints. The planning constraints set the preconditions that must be met before execution and the postconditions that need to be met after execution for each planning action.

[0028] Step S24: Update the scene graph: execute the planned action sequence while complying with the planning constraints, convert the initial scene graph into the current scene graph, generate feedback information of the planned action sequence, design an iterative planner, set an end mark, and execute the planned action sequence until the end mark is obtained.

[0029] Further, step S3 includes the following steps:

[0030] Step S31: Extract physical features: extract the road edges in the current scene graph and divide them into three categories: ordinary roads, main roads and expressways, collect the road surrounding configuration, traffic-related POIs, land use types and building outlines, and describe the physical features of the current scene graph from four dimensions. The physical features include population mobility, traffic facility construction, road topology, and spatial richness;

[0031] Step S32: Projecting the scene graph: constructing a probabilistic encoder and a deterministic decoder, projecting the current scene graph of the national land space planning to the feature dimension through a multi-layer perceptron, initializing N query subsets, and capturing spatial layout context information;

[0032] Step S33: constructing a planning evaluation model: the planning evaluation model includes a priori encoder, a posterior encoder, a decoder, a re-encoder, a convolutional encoder and a planner, the priori encoder, the posterior encoder, the decoder and the re-encoder include a transformer layer, a repeated self-attention block, a cross-attention block and a multi-layer perceptron, the planner includes two cross-attention blocks, and the planning evaluation model is used to comprehensively consider the spatial layout context information provided by the current scene graph;

[0033] Step S34: Calculate the parameters of the posterior distribution: Use the posterior encoder to calculate the parameters of the posterior distribution, and output the planning evaluation information through an additional cross-attention block;

[0034] Step S35: Calculate the prior distribution parameters: Use the prior encoder through the transformer layer to set the initial hypothesis, use the initial hypothesis to iteratively adjust the spatial layout context information for noise detection, capture the occlusion between objects and the repeated detection relationship of the same object, input the residual of the initial hypothesis into the multi-layer perceptron, estimate the mean vector and diagonal covariance matrix of the prior distribution, calculate the parameters of the prior distribution, and describe the planning evaluation information error;

[0035] Step S36: feature information fusion: use the decoder to generate a set of proxy bounding boxes for the planning evaluation information error, and perform feature information fusion on the initial hypothesis and the planning evaluation information error by repeating the self-attention block and the cross-attention block to form a fused feature;

[0036] Step S37: Generate label distribution: Use the re-encoder to re-encode the fused features to form encoded features, and at the same time obtain the physical features of the current scene graph from the convolutional encoder. Use the planner to focus on the encoded features and the physical features through two cross-attention blocks to obtain the label distribution of the current scene graph;

[0037] Step S38: Optimize the planning evaluation model: determine the planning evaluation index according to the label distribution of the current scene graph, further calculate the spatial planning loss, train the optimized planning evaluation model, and update the spatial layout context information in combination with the physical characteristics of the current scene graph.

[0038] Further, step S38 includes the following steps:

[0039] Step S381: Determine planning evaluation indicators: Use K-means clustering algorithm to aggregate the label distribution of the current scene graph, extract typical scenes for ordinary roads, main roads and expressways respectively, and determine the scene label of each typical scene, and use the scene label as the planning evaluation indicator;

[0040] Step S382: Calculate the spatial planning loss: Add a priori regularization incentive to the planning evaluation index, and combine the prior distribution to infer the prior distribution of the current scene graph through the probability encoder, and calculate the spatial planning loss. The formula used is as follows:

[0041] ;

[0042] in, Indicates the loss of space planning, represents the strength coefficient controlling the prior regularization excitation, represents the number of hypotheses for the treatment objects, n represents, z represents the latent variable, represents the prior distribution, represents a set of latent variables for planning errors, S represents a set of vectors describing N objects in the current scene graph, M represents the current rasterized scene graph, represents the land use coefficient, i and j represent the sequence number of the current scene graph, It represents the relative proportion of each planned building type in the i-th current scene graph. represents the economic development equilibrium coefficient, represents the population factor in the i-th current scene graph, represents the traffic coefficient, represents the intensity of economic activity in the jth current scenario graph, represents the distance attenuation coefficient, Represents the traffic impedance between the i-th current scene graph and the j-th current scene graph after distance attenuation;

[0043] Step S383: Optimize the planning evaluation model: Use the consistency distillation method to train the planning evaluation model. Given the current scene graph, generate noise samples by adding noise and discretizing it, perform backpropagation training to minimize the mean square distance, add sinusoidal position information embedding combined with the physical features of the current scene graph, and update the spatial layout context information.

[0044] The beneficial effects achieved by the present invention are as follows:

[0045] (1) The method described above integrates multi-source data, providing a rich, accurate and interrelated data foundation for national land space planning. The integration of different types of data comprehensively reflects the current situation and goals of the planning area, making planning decisions more scientific. Combining land use types with population distribution and economic development indicators helps to reasonably determine the functional positioning of different regions and achieve optimal allocation of land resources.

[0046] (2) The method constructs a scene graph structure and defines a relationship set to represent basic spatial planning relationships. The initial scene graph is generated by a scene graph generator, and a planning action sequence and constraints are generated. The iterative planner is used to execute the action sequence under the premise of complying with the constraints to update the scene graph. The various elements and their relationships in the national land space planning are intuitively presented in a graph structure, which is convenient for analyzing and understanding complex spatial layouts. The dynamic update mechanism makes the planning process visual and traceable, and can timely discover problems and conflicts in planning decisions. In urban renewal planning, the relationship between buildings and surrounding transportation and public service facilities is clearly displayed. Through the execution and feedback of the planning action sequence, planning decisions such as building layout adjustment and traffic route optimization are optimized, thereby improving the rationality and feasibility of planning.

[0047] (3) The method projects the scene graph into the feature dimension through a multi-layer perceptron, and uses the components to work together to calculate the posterior distribution parameters and the prior distribution parameters, and performs feature information fusion and label distribution generation. It can deeply mine the spatial layout context information in the scene graph, accurately capture the complex relationships and features between objects, and better consider the relationship between road topology, transportation facility construction and population mobility, thereby providing more accurate evaluation and optimization suggestions for spatial planning. Combined with features such as land use type and building outline, it can more reasonably judge the rationality and development trend of land use, which is helpful to formulate scientific land use planning strategies.

[0048] (4) The method takes into account various objectives and constraints in national land space planning more comprehensively, avoiding a single factor dominating the optimization direction of the model. The introduction of the land use coefficient prompts the model to focus on the rationality of the land use structure. The use of consistency distillation method and noise processing improves the robustness and generalization ability of the model, enabling it to better adapt to different planning scenarios and data changes, and provide more reliable evaluation and decision-making support for national land space planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 A flowchart of a national land space planning assessment method based on deep learning proposed by the present invention;

[0050] Figure 2 This is a flow chart of step S3 proposed by the present invention. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0052] Example 1: Reference Figure 1 This embodiment provides a land space planning assessment method based on deep learning, which specifically includes the following steps:

[0053] Step S1: data collection and preprocessing: collecting planning data of national land space planning, performing preprocessing operations on the planning data of national land space planning, obtaining preprocessed data, and generating a global dictionary;

[0054] Step S2: scene graph construction and update: construct the initial scene graph, generate the planned action sequence and planning constraints, execute the planned action sequence in compliance with the planning constraints, convert the initial scene graph into the current scene graph, and generate feedback information of the planned action sequence;

[0055] Step S3: Planning evaluation model deployment: extract the physical features of the current scene graph, build a planning evaluation model, obtain the label distribution of the current scene graph according to the physical features, determine the planning evaluation index, and further calculate the spatial planning loss. By minimizing the spatial planning loss, train and optimize the planning evaluation model;

[0056] Step S4: Spatial planning assessment application: Use the planning assessment model to conduct land use assessment, traffic condition assessment and socio-economic assessment, and generate an assessment result report.

[0057] Embodiment 2: This embodiment is based on the above embodiment, and step S1 includes the following steps:

[0058] Step S11: Collecting geospatial data: Collecting high-resolution satellite images, and using satellite images to provide land use type information and topographic information. The land use type information includes construction land, agricultural land and ecological land, and the topographic information includes altitude, slope and slope direction;

[0059] Step S12: Collecting social and economic data: Collecting population distribution data, economic development indicators and transportation network distribution information, where the economic development indicators include GDP and industrial structure data, and the transportation network distribution information includes flow data;

[0060] Step S13: Collect planning document data: collect planning objectives and spatial layout plans of national land space planning. Planning objectives include the proportion of ecological protection red lines and construction land scale control targets. Spatial layout plans include functional zoning and industrial layout.

[0061] Step S14: Preprocessing geospatial data: using ground control points, correcting the image to the correct geographic coordinate system through a polynomial fitting method, normalizing the geospatial data, and mapping the value range to a specific interval [0,1];

[0062] Step S15: preprocessing social and economic data: using the K-nearest neighbor algorithm to fill in missing values, for example, for economic data, convert it into a standard normal distribution with a mean of 0 and a standard deviation of 1 to eliminate the impact of dimensions between different indicators;

[0063] Step S16: Preprocessing planning file data: extracting key indicators and time nodes in planning objectives through natural language processing technology, and vectorizing the spatial layout plan;

[0064] Step S17: Construct a global dictionary: merge the names extracted from the relevant data of the national land space planning to form an initial name set, and delete duplicate names to generate a global dictionary.

[0065] Embodiment 3: This embodiment is based on the above embodiment, and step S15 includes the following steps:

[0066] Step S151: Processing missing values: For the missing flow data of certain road sections in the traffic network distribution information, the K-nearest neighbor algorithm is used to predict the missing flow data values ​​based on the flow data and flow data of surrounding road sections;

[0067] Step S152: Normalization processing: converting the data into a standard normal distribution form;

[0068] Step S153: Multi-source data fusion: Population distribution data, economic development indicators and transportation network distribution information are fused through weighted summation and feature splicing.

[0069] Embodiment 4: This embodiment is based on the above embodiment, and step S2 includes the following steps:

[0070] Step S21: define the scene graph: construct a planning scene graph structure G={V,E} based on the preprocessed data, where V is a node representing an object in the planning scene, and E is an edge describing the relationship between objects. The relationship set is constructed as R, representing the basic spatial planning relationship;

[0071] Step S22: Generate a scene graph: Use a scene graph generator to receive preprocessed data, a relationship set and a global dictionary according to the planned scene graph structure, and generate an initial scene graph, wherein the initial scene graph includes 128 objects;

[0072] Step S23: Define the planning task: According to the initial scene graph, design a spatial planner based on Transformer technology to generate a planning action sequence and planning constraints. The planning constraints set the preconditions that must be met before execution and the postconditions that need to be met after execution for each planning action.

[0073] Step S24: Update the scene graph: execute the planned action sequence while complying with the planning constraints, convert the initial scene graph into the current scene graph, and generate feedback information of the planned action sequence, design an iterative planner, receive the initial scene graph, the current scene graph, the planning constraints, and the feedback information of the planned action sequence, set an end mark, execute the planned action sequence until the end mark is obtained, and obtain the current scene graph, which reflects the spatial layout context information;

[0074] If the planned action violates the planning constraints, the error count is increased and feedback information is output until the end mark is obtained;

[0075] If the planning action passes the planning constraint verification, the current scene graph state is updated to reflect the execution of each planning action. For the object removal operation, the edge representing the initial support relationship is removed, and a new edge representing the new support relationship is created. The current scene graph is modified accordingly, and the next planning action in the planning action sequence is verified until the end mark is obtained, and the current scene graph is obtained to reflect the spatial layout context information.

[0076] Example 5: Reference Figure 2 This embodiment is based on the above embodiment, and step S3 includes the following steps:

[0077] Step S31: Extract physical features: extract the road edges in the current scene graph and divide them into three categories: ordinary roads, main roads and expressways, collect the road surrounding configuration, traffic-related POIs, land use types and building outlines, and describe the physical features of the current scene graph from four dimensions. The physical features include population mobility, traffic facility construction, road topology, and spatial richness;

[0078] Transportation-related POIs include subway stations, bus stops, and parking lots;

[0079] Land use types include green areas, administrative and public services, commercial and business facilities, educational areas, and industrial areas;

[0080] Building outlines include shape and height;

[0081] Road topology includes road betweenness centrality, spatial integration, angle choice, and intersection density;

[0082] Spatial richness includes land use types and building outlines;

[0083] Step S32: Projecting scene graph: constructing a probabilistic encoder and a deterministic decoder, projecting the current scene graph of the national land space planning to the feature dimension through a multi-layer perceptron, initializing 128 query subsets, and capturing spatial layout context information;

[0084] Step S33: constructing a planning evaluation model: the planning evaluation model includes a priori encoder, a posterior encoder, a decoder, a re-encoder, a convolutional encoder and a planner, the priori encoder, the posterior encoder, the decoder and the re-encoder include a transformer layer, a repeated self-attention block, a cross-attention block and a multi-layer perceptron, the planner includes two cross-attention blocks, and the planning evaluation model is used to comprehensively consider the spatial layout context information provided by the current scene graph;

[0085] Step S34: Calculate the posterior distribution parameters: Use the posterior encoder to calculate the parameters of the posterior distribution, output the planning evaluation information through an additional cross-attention block, and obtain the spatial detection features;

[0086] Step S35: Calculate the prior distribution parameters: Use the prior encoder through the transformer layer to set the initial hypothesis, use the initial hypothesis to iteratively adjust the spatial layout context information for noise detection, capture the occlusion between objects and the repeated detection relationship of the same object, input the residual of the initial hypothesis into the multi-layer perceptron, estimate the mean vector and diagonal covariance matrix of the prior distribution, calculate the parameters of the prior distribution, and describe the planning evaluation information error;

[0087] Step S36: feature information fusion: use the decoder to generate a set of proxy bounding boxes for the planning evaluation information error, and perform feature information fusion on the initial hypothesis and the planning evaluation information error by repeating the self-attention block and the cross-attention block to form a fused feature;

[0088] Step S37: Generate label distribution: Use the re-encoder to re-encode the fused features to form encoded features, and at the same time obtain the physical features of the current scene graph from the convolutional encoder, and use the planner to focus on the encoded features and the physical features through two cross-attention blocks to obtain the label distribution of the current scene graph;

[0089] Step S38: Optimize the planning evaluation model: determine the planning evaluation index according to the label distribution of the current scene graph, further calculate the spatial planning loss, train the optimized planning evaluation model, and update the spatial layout context information in combination with the physical characteristics of the current scene graph.

[0090] Embodiment 6: This embodiment is based on the above embodiment, and step S38 includes the following steps:

[0091] Step S381: Determine planning evaluation indicators: Use K-means clustering algorithm to aggregate the label distribution of the current scene graph, extract 4, 3, and 3 typical scenes for ordinary roads, main roads, and expressways, respectively, and determine the top 15 scene labels for each typical scene, and use the scene labels as planning evaluation indicators;

[0092] Step S382: Calculate the spatial planning loss: Add a priori regularization incentive to the planning evaluation index, and combine the prior distribution to infer the prior distribution of the current scene graph through the probability encoder, and calculate the spatial planning loss. The formula used is as follows:

[0093] ;

[0094] in, Indicates the loss of space planning, represents the strength coefficient controlling the prior regularization excitation, represents the number of hypotheses for the treatment objects, n represents, z represents the latent variable, represents the prior distribution, represents a set of latent variables for planning errors, S represents a set of vectors describing N objects in the current scene graph, M represents the current rasterized scene graph, represents the land use coefficient, i and j represent the sequence number of the current scene graph, It represents the relative proportion of each planned building type in the i-th current scene graph. represents the economic development equilibrium coefficient, represents the population factor in the i-th current scene graph, represents the traffic coefficient, represents the intensity of economic activity in the jth current scenario graph, represents the distance attenuation coefficient, Represents the traffic impedance between the i-th current scene graph and the j-th current scene graph after distance attenuation;

[0095] Step S383: Optimize the planning evaluation model: Initialized to the mean, the planning evaluation model is trained using the consistency distillation method. Given the current scene graph data points, noise samples are generated by adding noise and discretizing them. Backpropagation training is performed to minimize the mean square distance. Sinusoidal position information is added to embed and combined with the physical features of the current scene graph to update the spatial layout context information.

[0096] The present invention and its implementation methods are described above. Such description is not restrictive. If a person skilled in the art is inspired by it and creatively designs structures and implementation methods similar to the technical solution without departing from the purpose of the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A national land space planning evaluation method based on deep learning, characterized in that: The following steps are involved: Step S1: data collection and preprocessing: collecting planning data of national land space planning, performing preprocessing operations on the planning data of national land space planning, obtaining preprocessed data, and generating a global dictionary; Step S2: scene graph construction and update: construct the initial scene graph, generate the planned action sequence and planning constraints, execute the planned action sequence in compliance with the planning constraints, convert the initial scene graph into the current scene graph, and generate feedback information of the planned action sequence; Step S3: Planning evaluation model deployment: extract the physical features of the current scene graph, build a planning evaluation model, obtain the label distribution of the current scene graph according to the physical features, determine the planning evaluation index, calculate the spatial planning loss, and train and optimize the planning evaluation model by minimizing the spatial planning loss; Step S4: Spatial planning assessment application: Use the planning assessment model to perform assessment and generate an assessment result report.

2. According to claim 1, a method for evaluating national land space planning based on deep learning is characterized in that: Step S2 includes the following steps: Step S21: define the scene graph: construct the planning scene graph structure and relationship set based on the preprocessed data; Step S22: Generate a scene graph: Use a scene graph generator to receive preprocessed data, a relationship set and a global dictionary according to the planned scene graph structure, and generate an initial scene graph; Step S23: define the planning task: according to the initial scene graph, design a spatial planner based on the Transformer technology, generate a planning action sequence and planning constraints, and the planning constraints set preconditions and postconditions for each planning action; Step S24: Update the scene graph: execute the planned action sequence in accordance with the planning constraints, convert the initial scene graph into the current scene graph, generate feedback information of the planned action sequence, set the end mark, execute the planned action sequence until the end mark is obtained, and obtain the current scene graph.

3. The method for evaluating national land space planning based on deep learning according to claim 2 is characterized in that: Step S3 includes the following steps: Step S31: extracting physical features: extracting road edges in the current scene graph and describing the physical features of the current scene graph from four dimensions; Step S32: Projecting the scene graph: constructing a probabilistic encoder and a deterministic decoder, projecting the current scene graph to the feature dimension through a multi-layer perceptron, initializing the query subset, and capturing the spatial layout context information; Step S33: constructing a planning evaluation model: using the planning evaluation model to comprehensively consider the spatial layout context information provided by the current scene graph; Step S34: Calculate the posterior distribution parameters: Use the posterior encoder to calculate the parameters of the posterior distribution, output the planning evaluation information through an additional cross-attention block, and obtain the spatial detection features; Step S35: Calculate the parameters of the prior distribution: Use the prior encoder through the transformer layer to set the initial hypothesis, use the initial hypothesis to iteratively adjust the spatial layout context information for noise detection, input the residual of the initial hypothesis into the multi-layer perceptron, estimate the mean vector and diagonal covariance matrix of the prior distribution, calculate the parameters of the prior distribution, and describe the planning evaluation information error; Step S36: feature information fusion: using the decoder to perform feature information fusion on the initial hypothesis and planning evaluation information error by repeating the self-attention block and the cross-attention block to form a fused feature; Step S37: Generate label distribution: Use the re-encoder to re-encode the fused features to form encoded features, and at the same time obtain the physical features of the current scene graph from the convolutional encoder, and use the planner to focus on the encoded features and the physical features through two cross-attention blocks to obtain the label distribution of the current scene graph; Step S38: Optimize the planning evaluation model: determine the planning evaluation index according to the label distribution of the current scene graph, calculate the spatial planning loss, train the optimized planning evaluation model, and update the spatial layout context information.

4. The method for evaluating national land space planning based on deep learning according to claim 3 is characterized in that: Step S38 includes the following steps: Step S381: Determine planning evaluation indicators: Aggregate the label distribution of the current scene graph, extract typical scenes for the road side, determine the scene label of each typical scene, and use the scene label as a planning evaluation indicator; Step S382: Calculate the spatial planning loss: add a priori regularization encouragement to the planning evaluation index, and calculate the spatial planning loss in combination with the prior distribution; Step S383: Optimize the planning evaluation model: Use the consistency distillation method to train the planning evaluation model. Given the current scene graph, generate noise samples by adding noise and discretizing it, perform back propagation training, add sinusoidal position information embedding combined with the physical features of the current scene graph, and update the spatial layout context information.

5. The method for evaluating national land space planning based on deep learning according to claim 3 is characterized in that: The planning evaluation model includes a priori encoder, a posterior encoder, a decoder, a re-encoder, a convolutional encoder and a planner. The priori encoder, a posterior encoder, a decoder and a re-encoder contain transformer layers, repeated self-attention blocks, criss-cross attention blocks and multi-layer perceptrons. The planner includes two criss-cross attention blocks.

Images