Land reclamation collaborative planning system based on GIS and BIM fusion

Abstract

The invention relates to the technical field of land reclamation planning, and discloses a land reclamation collaborative planning system based on GIS and BIM fusion. The system comprises a basic data acquisition module which is used for acquiring a historical land GIS spatial data set, a historical BIM component data set and a historical improvement project data set of a target area; the multi-level integration module is used for carrying out space-time correlation integration on the data set to generate a land reclamation multi-source fusion data pool; the strategy generation module is used for performing multi-factor dynamic indexing in the data pool based on a preset renovation target and generating an initial collaborative planning strategy set; the strategy adjusting module is used for dynamically adjusting the initial strategy through a dual-channel regulation and control mechanism and outputting a target collaborative planning strategy; the execution feedback module drives planning execution and collects a real-time land state feedback data set; and the data pool updating module is used for comparing the difference factors of the feedback data and the preset target and updating the multi-source fusion data pool. The system improves the dynamics and accuracy of land reclamation planning.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of land consolidation planning, in particular to a land consolidation collaborative planning system based on GIS and BIM fusion. BACKGROUND

[0002] In the process of land resource development and utilization, land consolidation, as an important means of optimizing the layout of land space and improving the efficiency of land use, directly affects the sustainable development of the region. The planning of traditional land consolidation relies on a single data source, which often has problems such as limited data dimensions and insufficient analysis depth. GIS technology has advantages in spatial data management and geographic analysis, and can present the macro spatial distribution characteristics of regional land, but it is difficult to accurately depict the internal component properties and structural details of the land use unit. BIM technology is good at building three-dimensional models and attribute information of micro components, and can accurately describe the parameter characteristics of buildings, infrastructure and other entities, but it has limitations in macro spatial correlation and regional scale analysis. With the increasing diversification of land consolidation needs, a single technical means cannot meet the planning requirements in complex scenarios. In the current planning process, historical land space data, component detail data and consolidation project data are often stored in different platforms, and the data formats are not unified and the spatial and temporal references are inconsistent, which makes it difficult to effectively associate multi-source data. The generation of planning strategies often relies on experience and judgment or static model analysis, and lacks dynamic consideration of topography, resource distribution, ecological constraints and other factors, which may cause the strategies to deviate from the actual needs. At the same time, the real-time state feedback mechanism in the planning execution process is not perfect, and the deviation between the consolidation effect and the preset target is difficult to capture in time, and the existing data resources cannot be dynamically updated according to the actual changes, which leads to a lack of accurate data support for subsequent planning, and problems such as repeated consolidation and resource waste often occur. In the collaborative planning level, the data sharing of each participant is not smooth in the traditional mode, and there are information barriers in the planning, design and implementation links, making it difficult to form an efficient collaborative workflow. With the advancement of new urbanization construction and the improvement of ecological protection requirements, land consolidation faces more complex constraints, and how to achieve the organic connection between macro spatial planning and micro component design, and how to generate adaptive planning strategies based on dynamically changing data, have become problems that need to be solved in the field of land consolidation planning. SUMMARY

[0003] The purpose of the present application is to provide a land consolidation collaborative planning system based on GIS and BIM fusion to solve the problems raised in the background art.

[0004] To achieve the above purpose, the present application provides a land consolidation collaborative planning system based on GIS and BIM fusion, which comprises: The basic data collection module is configured to acquire a historical land GIS spatial dataset, a historical BIM component dataset, and a historical regulation project dataset of a target region. The multi-level integration module is configured to perform spatio-temporal correlation integration on the historical land GIS spatial dataset, the historical BIM component dataset, and the historical regulation project dataset, and generate a land regulation multi-source fusion data pool. The strategy generation module is configured to perform multi-factor dynamic indexing in the land regulation multi-source fusion data pool based on a preset regulation target, and generate an initial collaborative planning strategy set. The strategy adjustment module is configured to perform dynamic adjustment on the initial collaborative planning strategy set through a double-channel regulation mechanism, and output a target collaborative planning strategy. The execution feedback module is configured to drive planning execution according to the target collaborative planning strategy, and acquire a real-time land state feedback dataset. The data pool updating module is configured to compare difference factors between the real-time land state feedback dataset and the preset regulation target, and update the land regulation multi-source fusion data pool.

[0005] Preferably, the multi-level integration module is further configured to: extract vector boundary data of the historical land GIS spatial dataset, and perform spatial registration on the historical BIM component dataset; clean abnormal project records in the historical regulation project dataset based on a density clustering algorithm, and obtain a cleaned historical project set; associate the registered historical BIM component dataset and the cleaned historical project set to the vector boundary data, and construct a topological relationship of the land regulation multi-source fusion data pool.

[0006] Preferably, the strategy generation module is further configured to: define a spatial factor, an ecological factor, and an engineering factor as an indexing dimension, and construct a multi-factor dynamic indexing space; map the land regulation multi-source fusion data pool to the multi-factor dynamic indexing space, and generate a spatial indexing point set; perform a dynamic raster indexing operation in the spatial indexing point set with the preset regulation target as a constraint condition, and output the initial collaborative planning strategy set.

[0007] Preferably, the multi-level integration module is further configured to: identify a project confidence and an implementation effectiveness bias value in the cleaned historical project set; divide the cleaned historical project set into an expert experience layer, a key strategy layer, and a basic information layer according to the project confidence and the implementation effectiveness bias value; Integrate the layered cleaning history item set into the land consolidation multi-source fusion data pool.

[0008] Preferably, the strategy adjustment module is further configured to: Monitor an environmental change indicator of the real-time land state feedback dataset; If the environmental change indicator exceeds a preset threshold, activate the experience channel in the double-channel regulation mechanism, and extract a matching strategy from the expert experience layer; Otherwise, activate the intelligent channel, and optimize the initial collaborative planning strategy set using a reinforcement learning algorithm; Fuse the matching strategy or the optimized initial collaborative planning strategy set to generate the target collaborative planning strategy.

[0009] Preferably, the strategy generation module is further configured to: Select a reference strategy point in the key strategy layer; Calculate the feature offset of other strategy points in the key strategy layer from the reference strategy point; Use a sliding window time series sampling algorithm to reduce the dimension of the feature offset and generate a strategy feature sequence; Input the strategy feature sequence into the dynamic grid index operation.

[0010] Preferably, the strategy adjustment module is further configured to: Analyze the time series change pattern of the strategy feature sequence; Calculate the strategy evolution rate and the environmental response rate by a weighted moving average algorithm; Use the strategy evolution rate and the environmental response rate as optimization weight parameters of the reinforcement learning algorithm.

[0011] Preferably, the strategy adjustment module is further configured to: Obtain the current context features of the real-time land state feedback dataset; Match the current context features with the standard context templates of the basic information layer; Correct the strategy evolution rate and the environmental response rate according to the matching deviation value to generate a corrected evolution rate and a corrected response rate; Execute the reinforcement learning algorithm based on the corrected evolution rate and the corrected response rate.

[0012] Preferably, the execution feedback module is further configured to: Construct a strategy decision plane with the corrected evolution rate as the horizontal axis and the corrected response rate as the vertical axis; Divide the strategy decision plane into a low intervention zone, a medium intervention zone, and a high intervention zone; According to the landing area of the target collaborative planning strategy on the strategy decision plane, a planning execution instruction of a corresponding level is triggered.

[0013] Preferably, the data pool updating module is further configured to: extract actual engineering progress values and ecological index values of the real-time land state feedback data set; calculate a progress deviation factor of the actual engineering progress values from a preset engineering target and an ecological deviation factor of the ecological index values from a preset ecological standard; locate a to-be-updated level of the land consolidation multi-source fusion data pool according to the progress deviation factor and the ecological deviation factor; update historical project records in the to-be-updated level based on a gradient attenuation algorithm.

[0014] Compared with the prior art, the beneficial effects of the present application are: The land consolidation collaborative planning system based on GIS and BIM fusion breaks through the limitations of data isolation, static strategy and feedback lag in traditional land consolidation planning through multi-module collaborative operation. The basic data collection module includes historical land GIS spatial data set, historical BIM component data set and historical consolidation project data set in the unified collection range, changes the situation of single data source supporting planning in the past, so that the macro spatial information and micro component details required for planning can be fully integrated, and the planning deviation caused by incomplete data is avoided. The multi-level integration module integrates multi-source data in space and time, generates a land consolidation multi-source fusion data pool, and solves the problems of inconsistent data format and inconsistent space-time reference of different types of data. By establishing the space-time correlation relationship between the data, the macro spatial analysis capability of GIS and the micro component description capability of BIM are complementary, so that the planner can not only grasp the overall spatial layout of the regional land, but also deeply understand the component attributes in the specific plot, and provide more comprehensive data support for planning decision. The strategy generation module dynamically indexes multi-factors in the multi-source fusion data pool based on the preset consolidation target, changes the way of generating strategies by relying on experience or static models in traditional planning. By comprehensively considering factors such as terrain, soil, ecology, existing infrastructure and other aspects, the initial collaborative planning strategy set generated by dynamically selecting the adaptive planning strategy is more suitable for actual consolidation requirements, and the situation that the strategy does not match the actual scene is reduced. The double-path regulation mechanism of the strategy adjustment module provides a flexible path for planning strategy optimization. Based on the initial strategy, the dynamic adjustment mechanism can respond to the actual changes in different consolidation stages, considering both the rigid constraints of the preset target and the flexible needs of the on-site sudden situation, so that the output target collaborative planning strategy is more adaptive and operable. The execution feedback module collects land state feedback data in real time during the planning execution process, changing the traditional planning mode of paying more attention to redesign and less attention to feedback. By capturing the changes in land state during the improvement process in real time, the differences between planning execution and preset goals can be found in time, providing a basis for strategy adjustment and ensuring that the improvement process always moves towards the preset goal. The data pool updating module continuously updates the multi-source fusion data pool by comparing the differences between real-time feedback data and preset targets, so that the data resources can be continuously optimized with the improvement practice. The updated data pool not only provides support for the subsequent adjustment of the current improvement project, but also accumulates experience data for the land improvement planning of other similar areas, forms a virtuous cycle of data-driven planning, and improves the overall scientificity and effectiveness of land improvement planning. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 The timing diagram of the land improvement collaborative planning system based on GIS and BIM fusion according to the present application; Figure 2 The workflow diagram of the multi-level integration module; Figure 3 The workflow diagram of the strategy adjustment module; Figure 4 The generation process diagram of the strategy feature sequence. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0017] Please refer to Figure 1 The present application provides a land improvement collaborative planning system based on GIS and BIM fusion, which comprises: The basic data acquisition module is used to obtain a historical land GIS spatial data set, a historical BIM component data set and a historical regulation project data set of a target area; the multi-level integration module is used to perform spatio-temporal correlation integration on the historical land GIS spatial data set, the historical BIM component data set and the historical regulation project data set, and generate a land regulation multi-source fusion data pool; the strategy generation module is used to perform multi-factor dynamic indexing in the land regulation multi-source fusion data pool based on a preset regulation target, and generate an initial collaborative planning strategy set; the strategy adjustment module is used to dynamically adjust the initial collaborative planning strategy set through a double-channel regulation mechanism, and output a target collaborative planning strategy; the execution feedback module is used to drive planning execution according to the target collaborative planning strategy, and collect a real-time land state feedback data set; and the data pool updating module is used to compare difference factors of the real-time land state feedback data set and the preset regulation target, and update the land regulation multi-source fusion data pool.

[0018] Embodiment 1: refer to Figure 2 The multi-level integration module performs a spatio-temporal correlation integration operation. When processing the historical land GIS spatial data set, vector boundary data stored in a geographic coordinate system is extracted. The data contains spatial contour information of the target area, described in a polygon geometry structure. For the historical BIM component data set, a spatial registration operation is performed. The operation uses an affine transformation algorithm for coordinate conversion, and aligns the building component model to the GIS vector boundary by selecting control points at feature locations with clear spatial positioning such as building corner points and road intersections. The control points are selected at feature locations with clear spatial positioning such as building corner points and road intersections. After coordinate conversion, the registered historical BIM component data set is output, and its coordinate system is consistent with the GIS vector boundary. The historical regulation project data set needs to be cleaned. The density clustering algorithm is used to identify and remove outlier records. The DBSCAN method is used for the density clustering algorithm, and the parameter setting is based on the distribution characteristics of the historical data of the target area. The neighborhood radius is referred to the average spatial distance of similar projects, and the minimum point threshold is set according to the project cluster size. This process identifies isolated records that lack spatio-temporal correlation with other projects, and generates a cleaned historical project set.

[0019] After the above data processing, the integration module constructs the topological relationship of the multi-source fusion data pool. The registered historical BIM component dataset is mapped to the corresponding position of the GIS vector boundary, and each component binds its geographic coordinates. The cleaned historical project set is associated with the vector boundary through the space-time stamp, and the time stamp records the project start and end date, and the space stamp records the implementation area coordinate range. All associated data establish topological relationships through graph structure, and the nodes in the graph represent three types of entities: space block nodes represent land units divided by GIS vector boundaries, component nodes represent independent building components, and project nodes represent improvement project records. Connection edges define the relationship between entities: spatial adjacent edges connect adjacent block nodes, containing edges connect block nodes with their internal component nodes, and execution edges connect project nodes with the space block nodes they act on. All nodes have additional attribute data, space block nodes contain land use type attributes, component nodes contain material attributes, and project nodes contain construction parameter attributes.

[0020] The strategy generation module performs multi-factor dynamic indexing based on the multi-source fusion data pool. Three types of index dimensions are defined: the spatial dimension includes terrain slope and land use type data, the slope data is derived from digital elevation model calculation values, and the land use type is coded according to national classification standards; the ecological dimension includes vegetation coverage and soil quality index, the coverage is inverted through remote sensing image normalized vegetation index, and the soil quality index is a comprehensive index of organic matter content and pH value; the engineering dimension includes construction cost and material durability score, the cost data is taken from project final accounts, and the durability score is graded according to the service life of material types in standard environment. After normalizing each dimension data to dimensionless value, a multi-dimensional vector is generated to form an index space. The number of vector dimensions is equal to the total number of defined index factors.

[0021] When mapping the multi-source fusion data pool to the index space, the feature extraction algorithm processes each valid project record in the pool. Feature extraction includes numerical conversion and vectorization: spatial factor extraction of slope mean value and main utilization type proportion, ecological factor calculation of vegetation coverage difference before and after implementation and soil quality change rate, and engineering factor extraction of unit area cost and material durability score. The converted feature values are combined into a multi-dimensional vector and stored as an index point in the space index point set. The point set stores the mapping relationship between the original project data and the vector features.

[0022] The dynamic grid index operation is performed on the index point set, with the predetermined rectification target as the filtering condition. The rectification target includes the engineering time range, the total amount of funds upper limit, and the ecological restoration index threshold. The index process adopts variable granularity grid division: the initial grid division refers to the index point density distribution, and the point set sparse area uses large-scale grid, and the dense area automatically subdivides the grid. The grid subdivision is dynamically triggered according to the local area index point standard deviation, and the standard deviation exceeding the set value triggers the recursive grid division to the minimum unit. All grid units are traversed, and the index points meeting the target constraints are screened: the time condition checks whether the project date is within the set range, the fund condition compares the cost with the budget upper limit, and the ecological condition verifies the restoration index compliance status. The screened point set corresponds to the initial constraint compliant project strategy, forming the initial collaborative planning strategy set.

[0023] The strategy feature sequence enhances the index dynamicity. The reference strategy point is selected in the key strategy layer, and the reference point selection criteria include project frequency or spatial coverage. The feature offset of other strategy points in the layer from the reference point is calculated, and the offset is reflected by the vector difference. The time dimension compression is performed on the offset sequence: the sliding window moves at a fixed step, and the window size is set according to the data sampling frequency. The offset data in the window is averaged to generate a simplified offset value representing the time period. The simplified value sequence is added to the original index point as a supplementary time feature. The complete output flow of the strategy generation module maintains the spatio-temporal correlation between data, and the multi-dimensional design of the index space supports comprehensive condition retrieval.

[0024] The data pool structure output by the multi-level integration module contains three logical layers: the expert experience layer stores high-confidence project strategies, the key strategy layer concentrates high-frequency implementation project cases, and the basic information layer retains basic environmental parameters. The hierarchical logic is based on project confidence and effectiveness bias value calculation, the confidence is obtained by weighting expert evaluation label and historical verification frequency, and the effectiveness bias value compares the difference between project expected target and actual result. The strategy generation module performs the core index operation in the key strategy layer, and the project records in this layer have typical implementation characteristics. The index point set generation process is associated with the original data attributes, supporting reverse tracing of data sources. The entire implementation process maintains data consistency and traceability, and the spatial registration accuracy affects the accuracy of data association, and the density clustering parameter setting balances the cleaning intensity and data integrity.

[0025] Example 2: see Figure 3The multi-level integration module performs data layering operations after completing the construction of the land consolidation multi-source fusion data pool. This operation processes and cleans the records in the historical project set, calculates the project confidence and implementation effectiveness bias value. The project confidence is quantified by double data sources: the historical data verification rate counts the frequency of repeated references to project records in similar areas, and the expert evaluation score comes from independent scoring of project feasibility by field experts. The two values are weighted and fused to generate the final confidence index. The implementation effectiveness bias value calculates the difference between the expected outcome and the actual result of the project. The expected outcome is extracted from the quantitative targets in the project planning document, and the actual result comes from the acceptance report data. The bias value is expressed in percentage form to represent the target achievement degree.

[0026] Based on the confidence and bias value feature vectors, a hierarchical clustering operation is performed. The K-means method is selected for the clustering algorithm, and three cluster centers are set to correspond to the hierarchical structure. After normalization, the feature vectors are input into the algorithm, and the Euclidean distance between each project point and the cluster center is iteratively calculated. The termination condition of clustering is set as the threshold value of the center movement distance or the maximum number of iterations. The output result divides the project records into three independent data sets: the expert experience layer contains high-confidence and low-bias projects, the key strategy layer contains medium-confidence and medium-bias projects, and the basic information layer contains low-confidence and high-bias projects. Each layer of data set is labeled with metadata to identify the hierarchical attribute.

[0027] The cleaned historical project set after layering is integrated into the land consolidation multi-source fusion data pool. The integration operation is realized through the hierarchical index field of the data pool, and each project record is associated with a hierarchical label. The data pool storage structure adopts a tree-like organization, with the top layer being a spatial block node, and the lower layers hanging the project record sets of the expert experience layer, the key strategy layer, and the basic information layer. Reference relationships are set between layers to allow cross-layer data access.

[0028] The strategy adjustment module continuously monitors the real-time land state feedback data set. The environmental change indicators include dynamically updated parameters such as precipitation frequency and soil erosion rate, and the data comes from the Internet of Things sensor network. The preset threshold values are set based on historical meteorological data and geological reports of the target area and are stored in an independent threshold value library. The monitoring period and data collection frequency are synchronized, and threshold comparison is performed every time a new data set is obtained.

[0029] The double-channel regulation mechanism is triggered according to the state of the environmental indicators. When the monitoring parameters exceed the preset threshold values, the experience channel is activated: the system extracts the current environmental indicator vector and calculates its similarity to the environmental features of the projects stored in the expert experience layer. The similarity algorithm uses the standardized Euclidean distance, and the smaller the distance value, the higher the matching degree. The expert experience layer project records are traversed, and the top K records with the smallest distance are selected as candidate strategies. The candidate strategies are filtered for timeliness, and only the effective solutions within the last five years are retained, and the matching strategy set is output.

[0030] When the environmental indicators are within the normal threshold range, the intelligent channel is activated: the system takes the initial set of collaborative planning strategies as input and optimizes them using a reinforcement learning algorithm. The algorithm constructs a Markov decision process model, with the state space defined as the strategy feature vector and the action space containing policy parameter adjustment operations. The reward function is designed to consider factors such as engineering efficiency, ecological impact, and economic cost. The Q-learning algorithm iteratively updates the Q-value table to explore the optimal policy adjustment path. The optimization process continues until the reward function converges or the maximum number of iterations is reached, and the optimized strategy set is output.

[0031] The target collaborative planning strategy is generated through a fusion mechanism. The matching strategy set output by the experience channel or the optimized strategy set output by the intelligent channel serves as the main input source. The fusion process uses a weighted average method: the experience channel strategy weight is assigned based on the confidence of the source project, and the intelligent channel strategy weight is determined based on the reward value when the reinforcement learning converges. After weight allocation, perform strategy parameter fusion calculation: take the weighted average value for the same type of parameters, and use the majority voting mechanism for discrete parameters. The fusion result generates the final target collaborative planning strategy document.

[0032] The strategy adjustment module maintains a record of dual-channel switching. Each time the channel is activated, the trigger reason, input parameters, and processing time are recorded. Historical switching data are stored in an independent log library for analyzing system decision patterns. The experience channel matching process retains intermediate values in similarity calculation, and the intelligent channel optimization process records Q-value table update trajectories. These process data support strategy backtracking verification.

[0033] In the data hierarchical operation, a time decay factor is introduced in the confidence calculation, giving higher weight to recent expert scores. An abnormal processing mechanism is set for hierarchical clustering: when the number of projects in a certain category is too small, the clustering center position is automatically adjusted. Priority rules are set for dual-channel activation: when multiple indicators exceed the limit at the same time, they are processed in order of severity. The matching strategy extraction process includes conflict detection, which coordinates candidate strategies with conflicting parameters. The reinforcement learning model sets an exploration rate decay mechanism: the probability of random exploration gradually decreases as the number of iterations increases. The fusion calculation process includes a parameter normalization step to eliminate the influence of different dimensions. The entire implementation process responds to data updates through an event-driven mechanism, and the output results of each link are written in real time to the shared storage area.

[0034] Example 3: see Figure 4The strategy generation module performs feature processing within the key strategy layer. Benchmark strategy points are selected using a dynamic frequency weighting method, calculating a comprehensive score for each project record across the dimensions of spatial coverage, implementation frequency, and number of associated components. A time decay factor is introduced into the score calculation, giving higher weight to recent projects. The system automatically selects the highest-scoring record as the benchmark point. In the event of a tie, the project with the highest spatial coverage is prioritized. Once the benchmark point is determined, all other strategy points within that layer are traversed, calculating the characteristic offset of each point relative to the benchmark. The characteristic offset is calculated based on the difference in spatial, ecological, and engineering factors: the spatial offset reflects the difference in terrain slope and the degree of change in land use type; the ecological offset quantifies the difference in vegetation cover change and soil quality change; and the engineering offset expresses the difference in cost deviation and material durability. Each offset result is stored as a three-dimensional vector, recording the relative deviation between the original strategy point and the benchmark point in the corresponding dimension.

[0035] After the offset sequence is generated, time series compression is performed. The system loads the offset vector sequence sorted by timestamp within the historical implementation period and uses the sliding window time series sampling algorithm for dimensionality reduction. Set a fixed length time window The number of consecutive time points covered by the window is determined by the data sampling frequency. Slide along the time axis, covering new data points with each movement. Within a single window, perform mean calculations on the multiple offset vectors contained therein. Specific operations are processed separately according to the spatial, ecological, and engineering dimensions: in the spatial dimension, the arithmetic mean of the slope differences within the window is taken, and the mode of land use type change is taken; in the ecological dimension, the sliding mean of the vegetation change rate is calculated, and the median of the soil quality change amplitude is taken; in the engineering dimension, the weighted average of the cost deviation rate is calculated, and the maximum value of the material durability difference is taken. This processing process can be formally expressed as the following unique formula:

[0036] in: represents the compressed feature vector output by the kth window; is the preset window length (time unit); Represents the feature offset at the original time point i. The summation range is the time point sequence covered by the current window. Each window outputs a simplified feature vector, forming a reduced-dimensional strategy feature sequence. This sequence retains the trend characteristics of the original data while significantly reducing the data size.

[0037] The strategy feature sequence is input into a dynamic raster indexing operation. The system uses each vector in the sequence as a supplementary attribute in the time dimension of the spatiotemporal index point. During the indexing operation, the spatial location information remains in its original GIS coordinates, but the feature attributes are updated to the sequence values. This mechanism enables the indexing process to reflect the dynamic evolution of strategy features.

[0038] The strategy adjustment module analyzes the time series characteristics of the strategy feature sequence. The system uses Fourier transform to analyze the periodicity of the sequence and identify the main fluctuation frequency and its amplitude. The trend item decomposition uses a smoothing algorithm to separate the long-term trend from the short-term fluctuation and extract the basic evolution direction of the strategy feature. Based on the decomposition results, two key parameters are calculated by the weighted moving average algorithm: the strategy evolution rate represents the change rate of the strategy feature vector per unit time, and the environmental response rate quantifies the correlation between the environmental index change and the strategy feature adjustment.

[0039] The weighted moving average algorithm sets up two independent calculation channels. The strategy evolution rate channel inputs the time difference sequence of the feature vector, and the weight distribution uses an exponential decay mode, with recent data getting higher weight. The environmental response rate channel inputs the covariant sequence of the environmental sensor data and the feature vector, and the weight is set according to the importance classification of environmental events. The calculation results are output as the evolution rate scalar ER and the response rate vector RV, where the vector components correspond to different types of environmental factors.

[0040] The reinforcement learning algorithm receives the evolution rate ER and the response rate RV as optimization weight parameters. The algorithm incorporates the ER value into the state transition probability calculation, adjusting the evolution expectation of the strategy state. The response rate vector RV is decomposed into environmental factor weights, which are used to modify the environmental cost coefficients in the reward function. In the Q-learning update rule, the environmental-related reward term is multiplied by the corresponding response rate component, so that the algorithm optimization direction and environmental sensitivity remain synchronized.

[0041] In the feature offset calculation stage, the spatial dimension difference processing needs to consider the geographical constraints. When two strategy points are located in different geomorphic units, the system automatically activates the terrain correction coefficient. The engineering dimension cost bias calculation introduces the time calibration of material price index to eliminate the impact of inflation. The time series analysis process sets an outlier filtering mechanism, which truncates data points exceeding three times the standard deviation. The reinforcement learning parameter transfer adopts incremental update, and the evolution rate and response rate parameters perform rolling refresh after each environmental feedback data arrives. The window sampling algorithm sets edge data processing rules, and the start and end of the time series use asymmetric window padding. The whole implementation process enhances the dynamic adaptability of the time series feature index, and the weight parameter transfer optimizes the response characteristics of the decision model.

[0042] Example 4: The strategy adjustment module performs contextual adaptive optimization in the scenario of a saline-alkali land improvement project. The real-time land state feedback dataset comes from 12 monitoring nodes deployed in the project area, and the current collection cycle returns environmental parameters including: air temperature 31.2℃, air humidity 58%, soil electrical conductivity 4.8dS / m, groundwater level 1.2m, pH value 8.6. The system standardizes the original data into a feature vector [0.82, 0.61, 0.93, 0.35, 0.88]. The standard context template stored in the basic information layer contains six typical saline-alkali land environment modes, among which the "coastal medium salt type" template has a benchmark value of [air temperature 28℃, humidity 65%, electrical conductivity 4.5dS / m, water level 1.5m, pH 8.5].

[0043] The feature matching engine calculates the Manhattan distance of the current context from each template: the "coastal medium salt type" has the smallest distance value (total deviation 0.48). After the system determines this as the best matching template, it generates a dimensional deviation vector: air temperature +11.4%, humidity -10.8%, electrical conductivity +6.7%, water level -20.0%, pH +1.2%. This deviation vector triggers strategy parameter correction: the original strategy evolution rate 0.76 is adjusted by reducing 0.05 according to the negative water level deviation and increasing 0.03 according to the positive electrical conductivity deviation, generating a corrected evolution rate of 0.74. The environmental response rate original vector [0.18, 0.22, 0.35, 0.15, 0.10] corresponds to the five factor weights, and the negative humidity deviation reduces its weight by 0.04, while the positive electrical conductivity deviation increases its weight by 0.06, generating a corrected response rate of [0.18, 0.18, 0.41, 0.15, 0.10]. The reinforcement learning algorithm optimizes the drainage scheme using the corrected parameters, and strengthens the influence weight of the salt migration factor in the strategy evaluation.

[0044] The execution feedback module constructs the strategy decision plane, with the horizontal axis being the corrected evolution rate 0.74 and the vertical axis being the corrected response rate sum 1.02. After normalization of the coordinate system, it is mapped to the range [0, 1.5], and the current point coordinate is (0.74, 1.02). Based on the decision data clustering of 356 historical saline-alkali land projects, the plane is divided into three regions: low intervention area (coordinate sum ≤1.0), medium intervention area (1.0< coordinate sum ≤1.7), and high intervention area (coordinate sum >1.7). The current coordinate sum 1.76 falls into the high intervention area. The target collaborative planning strategy triggers a three-level execution instruction set, as shown in Table 1.

[0045] Table 1: High intervention instruction set for saline-alkali land projects.

[0046]

[0047] The partition threshold dynamic updating mechanism of the policy decision plane is activated every quarter. The system collects the decision points of the newly added 42 projects, and recalculates the boundaries through K-means clustering: the low intervention zone threshold is raised to 1.05, and the high intervention zone threshold is adjusted to 1.75. The current coordinate point is located in the buffer zone of the new threshold boundary (1.76 > 1.75), triggering the automatic review process. After reviewing and comparing historical data of similar projects, the high intervention determination is maintained.

[0048] After the execution instruction is issued, the mechanical scheduling system responds immediately: it marks the available underground pipe laying machine position on the equipment map, generates the optimal path to deploy 2 devices to the project area. The material allocation system starts the emergency procurement agreement, calculates the incremental demand for gypsum and sends an electronic order to the supplier. The environmental sensor network adjusts the monitoring protocol, adding water level sensors at three key profiles, and shortening the sampling frequency from every 6 hours to every 2 hours. The entire instruction execution process is recorded in real-time operation logs, including device start and stop time, material consumption, sensor readings, etc.

[0049] The context matching process sets the composite environment scenario processing rules. When the conductivity > 4.5 dS / m and the groundwater level < 1.3 m are monitored, the system automatically activates the "salt drought composite type" matching mode. The strategy parameter correction uses a bias gradient response mechanism: when the conductivity bias is in the 5%-10% interval, a 0.02 change in response rate is triggered for every 1% bias. The execution instruction set contains priority labels, and high-intervention instructions require the first round of operations to be started within 120 minutes. The strategy adjustment module receives new environmental data every 6 hours, and automatically downgrades the intervention level when the conductivity drops by more than 5% for two consecutive times.

[0050] In Example 5, when the data pool updating module starts the updating process, it first extracts the real-time land state feedback data set collected by the execution feedback module. This data set contains dynamic monitoring information during the implementation of the improvement project, involving two types of core parameters: actual project progress indicators and ecological monitoring indicators. Actual project progress indicators are sourced from the automated construction log system, recording the completion percentage, material consumption, and work hour statistics of each sub-project. Ecological monitoring indicators are collected by the sensor network deployed in the project area, including vegetation recovery index, soil and water conservation rate, and other quantitative values. These raw data are stored in the temporary buffer area after format conversion and time stamp alignment.

[0051] The process of calculating the deviation factor of the module includes two independent channels. The progress deviation channel processes the difference between the actual project progress value and the preset project target, which is derived from the quantitative indicators in the target collaborative planning strategy document. The difference calculation considers the time dimension and the completion dimension, and outputs the progress deviation factor. The ecological deviation channel compares the ecological index value with the preset ecological standard, which is set according to the regional ecological baseline. The comparison process distinguishes the weight of different ecological elements, and outputs the ecological deviation factor. Both deviation factors are converted into dimensionless percentage form and stored as updated decision parameters.

[0052] According to the combination of deviation factors, the level to be updated is located. The data pool update module maintains a level positioning matrix that defines the update trigger conditions of the expert experience layer, the key strategy layer, and the basic information layer. The positioning algorithm maps the progress deviation factor and the ecological deviation factor to the matrix coordinates, and determines the range of the level to be updated by threshold comparison. The expert experience layer corresponds to high deviation scenarios, the key strategy layer responds to medium deviation changes, and the basic information layer handles regular deviations. The positioning result outputs a set of identifiers of the level to be updated, including specific data levels that need to be adjusted.

[0053] The gradient attenuation algorithm is executed to update the historical project records. The system retrieves the set of project records in the level to be updated, each record being attached with a timestamp and an initial weight. The attenuation algorithm sets a weight adjustment rule based on the time distance, with recent records retaining higher weights and distant records gradually decreasing in weight. The weight update formula introduces a record age coefficient, which decreases exponentially with the increase of project age. The algorithm iterates through the target record set to recalculate the composite weight value of each record. The updated records are simultaneously modified in their storage weights in the data pool, while retaining the original data version for traceability.

[0054] The data pool structure reorganization operation is started after the weight update. The system detects the changes in the weight distribution of records in each level, and triggers structure optimization when the weight variance of a level exceeds a certain value. The optimization process recalculates the inter-level correlation strength and adjusts the level topology. For records with significantly reduced weights, the system moves them into the historical archive; for newly recorded records with significantly increased weights, they may be upgraded to higher levels of storage. Structure reorganization maintains the three-layer infrastructure of the data pool, but allows dynamic flow of records between levels.

[0055] The update process includes a data consistency check mechanism. After each modification of the record weight, the system automatically checks the logical consistency of the record with other associated data. When conflicts are found, a coordination process is started: engineering record conflicts are covered by the latest construction standards, and ecological record conflicts are calibrated by reference to authoritative monitoring data. The data that passes the verification generates a version snapshot, recording the update time, operator, and modification content.

[0056] The module sets the incremental update trigger condition. The system continuously monitors the rate of change of the deviation factor, and when the same direction growth appears for three consecutive collection periods, the update process is immediately started. For special cases of sudden deterioration of ecological indicators, an emergency update channel is set to skip the regular process. The update log records the deviation factor value, update level distribution and processing record quantity of each operation, forming an auditable operation track. The entire implementation process maintains the timeliness of the data pool through the deviation response mechanism, and the gradient decay ensures the reasonable decay rhythm of historical data.

[0057] The hierarchical positioning matrix regular calibration mechanism runs every half year. The calibration process analyzes the deviation pattern in the historical update record and adjusts the trigger threshold range of each level. The gradient decay algorithm introduces an environmental adaptation factor to automatically extend the decay period of important records in ecologically sensitive areas. The data pool update module and the strategy generation module establish a feedback link, and the updated data pool state notifies the strategy generation module in real time, triggering possible strategy optimization process.

[0058] It should be noted that, in the present document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0059] While embodiments of the present application have been shown and described with reference to particular embodiments thereof, it will be understood by those skilled in the art that various changes in form and details can be made therein without departing from the spirit and scope of the application. The scope of the application is defined by the appended claims and their equivalents.

Claims

1. A land consolidation collaborative planning system based on GIS and BIM fusion, characterized in that, The system comprises: a basic data acquisition module configured to acquire a historical land GIS spatial data set, a historical BIM component data set, and a historical regulation project data set of a target region; a multi-level integration module configured to perform spatio-temporal correlation integration on the historical land GIS spatial data set, the historical BIM component data set, and the historical regulation project data set, and generate a land regulation multi-source fusion data pool; a strategy generation module configured to perform multi-factor dynamic indexing in the land regulation multi-source fusion data pool based on a preset regulation target, and generate an initial collaborative planning strategy set; a strategy adjustment module configured to perform dynamic adjustment on the initial collaborative planning strategy set through a double-channel regulation mechanism, and output a target collaborative planning strategy; an execution feedback module configured to drive planning execution according to the target collaborative planning strategy, and acquire a real-time land state feedback data set; a data pool updating module configured to compare difference factors between the real-time land state feedback data set and the preset regulation target, and update the land regulation multi-source fusion data pool. 2.The GIS and BIM fusion-based land consolidation collaborative planning system according to claim 1, characterized in that, The multi-level integration module is further configured to: extract vector boundary data of the historical land GIS spatial data set, and perform spatial registration on the historical BIM component data set; clean abnormal project records in the historical regulation project data set based on a density clustering algorithm, and obtain a cleaned historical project set; associate the registered historical BIM component data set and the cleaned historical project set to the vector boundary data, and construct a topological relationship of the land regulation multi-source fusion data pool. 3.The land consolidation collaborative planning system based on GIS and BIM fusion according to claim 2, characterized in that, The strategy generation module is further configured to: define a spatial factor, an ecological factor, and an engineering factor as an indexing dimension, and construct a multi-factor dynamic indexing space; map the land regulation multi-source fusion data pool to the multi-factor dynamic indexing space, and generate a spatial indexing point set; perform a dynamic raster indexing operation in the spatial indexing point set with the preset regulation target as a constraint condition, and output the initial collaborative planning strategy set.

4. The land consolidation collaborative planning system based on GIS and BIM fusion of claim 3, wherein, The multi-level integration module is further configured to: identify a project confidence and an implementation effectiveness bias value in the cleaned historical project set; divide the cleaned historical project set into an expert experience layer, a key strategy layer, and a basic information layer according to the project confidence and the implementation effectiveness bias value; integrate the layered cleaned historical project set into the land regulation multi-source fusion data pool. 5.The land consolidation collaborative planning system based on GIS and BIM fusion according to claim 4, characterized in that, The strategy adjustment module is further configured to: monitor an environmental change index of the real-time land state feedback data set; if the environmental change index exceeds a preset threshold, activate an experience channel in the double-channel regulation mechanism, and extract a matching strategy from the expert experience layer; otherwise, activate an intelligent channel, and optimize the initial collaborative planning strategy set by using a reinforcement learning algorithm; fuse the matching strategy or the optimized initial collaborative planning strategy set, and generate the target collaborative planning strategy. 6.The land consolidation collaborative planning system based on GIS-BIM fusion of claim 5, wherein, The strategy generation module is further configured to: select a reference strategy point in the key strategy layer; calculate a feature offset of other strategy points in the key strategy layer from the reference strategy point; perform dimension reduction compression on the feature offset by using a sliding window time series sampling algorithm, and generate a strategy feature sequence; inputting the policy feature sequence into the dynamic grid index operation.

7. The land consolidation collaborative planning system based on GIS and BIM fusion of claim 6, wherein, The policy adjustment module is further configured to: analyze a time sequence change mode of the policy feature sequence; calculate a policy evolution rate and an environment response rate by a weighted moving average algorithm; use the policy evolution rate and the environment response rate as optimization weight parameters of the reinforcement learning algorithm. 8.The land consolidation collaborative planning system based on GIS-BIM fusion of claim 7, wherein, The policy adjustment module is further configured to: obtain a current context feature of the real-time land state feedback dataset; match the current context feature with a standard context template of the basic information layer; correct the policy evolution rate and the environment response rate according to a matching deviation value, to generate a corrected evolution rate and a corrected response rate; execute the reinforcement learning algorithm based on the corrected evolution rate and the corrected response rate. 9.The land consolidation collaborative planning system based on GIS-BIM fusion of claim 8, wherein, The execution feedback module is further configured to: construct a policy decision plane with the corrected evolution rate as the horizontal axis and the corrected response rate as the vertical axis; divide the policy decision plane into a low intervention zone, a medium intervention zone and a high intervention zone; trigger a planning execution instruction of a corresponding level according to a landing area of the target collaborative planning policy on the policy decision plane. 10.The GIS and BIM fusion-based land consolidation collaborative planning system according to claim 9, characterized in that, The data pool updating module is further configured to: extract an actual engineering progress value and an ecological index value of the real-time land state feedback dataset; calculate a progress deviation factor of the actual engineering progress value from a preset engineering target, and an ecological deviation factor of the ecological index value from a preset ecological standard; locate a to-be-updated level of the land consolidation multi-source fusion data pool according to the progress deviation factor and the ecological deviation factor; update a historical project record in the to-be-updated level based on a gradient attenuation algorithm.

Images