Construction land surveying and mapping information management method and system based on Internet of Things, and medium

Abstract

The invention relates to the technical field of information management, in particular to a construction land surveying and mapping information management method and system based on the Internet of Things and a medium. The method comprises the following steps: obtaining construction land surveying and mapping data through the Internet of Things, and carrying out collaborative perception fusion to obtain a construction land multi-mode perception matrix; performing multi-source consistency correction on the multi-modal sensing matrix of the construction land to obtain a standardized surveying and mapping tensor; obtaining soil quality data of the construction land, and analyzing a soil quality evolution trend of the construction land based on the soil quality data of the construction land; adjusting the semantic weight of the standardized surveying and mapping tensor according to the soil evolutionary trend of the construction land and the standardized surveying and mapping tensor to obtain a surveying and mapping enhanced semantic hypergraph; and performing land parcel type adaptive sparse coding on the surveying and mapping enhanced semantic hypergraph to obtain a real-time surveying and mapping flow data block, and uploading the real-time surveying and mapping flow data block to a land information management platform to perform an information storage task. According to the method, the data retrieval efficiency and accuracy of the surveying and mapping data can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of information management, and particularly to a method, a system and a medium for managing construction land surveying and mapping information based on the Internet of Things. Background Art

[0002] Traditional methods for managing construction land surveying and mapping information mainly rely on manual surveys and manual records, with low efficiency in information collection and processing, and large errors and uncertainties. Traditional construction land surveying and mapping management methods mainly include manual surveying, manual recording, and the application of traditional Geographic Information Systems (GIS), etc. Manual surveying requires a large amount of manpower, material resources and time. Especially in large-scale surveying tasks, it often relies on a large number of surveyors for on-site operations, which not only consumes a large amount of resources, but also has low measurement accuracy. Manual recording and paper file management also have problems such as information loss, difficulty in sharing, and high maintenance costs, which greatly restrict the timeliness and accuracy of data. On the other hand, although traditional GIS systems play a good role in data visualization and management in the management of construction land information, their limitations cannot be ignored. Traditional GIS is usually based on static data, lacks deep integration with real-time dynamic data, and is difficult to monitor and update construction land in real time. In addition, when dealing with large-scale surveying and mapping data, traditional GIS often has defects such as slow data processing speed, insufficient accuracy, and poor adaptability to complex terrains and environmental changes. For some key data such as land boundaries, geological conditions, and land uses, traditional systems cannot perform accurate analysis, affecting the decision-making support function of construction land planning. Summary of the Invention

[0003] Based on this, it is necessary for the present invention to provide a method, a system and a medium for managing construction land surveying and mapping information based on the Internet of Things to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for managing construction land surveying and mapping information based on the Internet of Things includes the following steps:

[0005] Step S1: Obtain construction land surveying and mapping data through the Internet of Things, and perform collaborative perception fusion on the construction land surveying and mapping data to obtain a construction land multi-modal perception matrix;

[0006] Step S2: Perform multi-supervised deviation removal on the construction land multi-modal perception matrix to obtain a construction land perception tensor, and perform multi-source consistency correction on the construction land perception tensor to obtain a standardized surveying and mapping tensor;

[0007] Step S3: Obtain construction land soil data, and analyze the evolution trend of construction land soil based on the construction land soil data;

[0008] Step S4: Conduct causal reasoning on the organic matter content based on the evolution trend of the construction land soil quality and the standardized mapping tensor to obtain the influencing factors of the geological organic matter content; adjust the semantic weights of the standardized mapping tensor based on the influencing factors of the geological organic matter content to obtain a mapping enhanced semantic hypergraph;

[0009] Step S5: Perform adaptive sparse coding of the plot types on the mapping enhanced semantic hypergraph to obtain real-time mapping stream data blocks, and upload them to the land information management platform for information storage tasks.

[0010] The present invention realizes the real-time acquisition and multi-modal fusion of construction land surveying and mapping data through Internet of Things technology, improving the breadth and accuracy of data acquisition. Compared with traditional manual surveying and mapping that relies on a single measurement method, resulting in limited data coverage and large errors, the present invention utilizes multi-sensor collaborative perception to fuse data from lidar, remote sensing images, and ground surveying and mapping equipment to construct a multi-modal perception matrix, thereby enhancing the comprehensiveness and reliability of surveying and mapping data and reducing measurement errors that may be brought by a single data source. In addition, through multi-supervised debiasing processing of surveying and mapping data, systematic biases caused by sensor measurement errors, environmental interference, and data noise are eliminated, improving the authenticity of surveying and mapping data. On this basis, a multi-source consistency correction technology is adopted to make different surveying and mapping data sources consistent in spatial coordinates, time synchronization, and data scale, solving the problem of information mismatch caused by inconsistent data sources in traditional GIS systems, and finally generating a standardized surveying and mapping tensor to support subsequent data analysis and modeling. Combining the analysis of soil quality data, the modeling and prediction of the evolution trend of construction land soil quality are realized. Traditional surveying and mapping management methods mainly rely on static soil quality data, which are difficult to capture the dynamic changes of soil properties and affect the rational planning of land use. The present invention establishes an analysis model for the evolution trend of soil quality by fusing data such as soil moisture, mineral composition, and nutrient content, enabling surveying and mapping data to not only be limited to topographic and geomorphic information, but also combine geological factors to provide a more accurate reference for construction land planning. Further, the causal reasoning method is used to analyze the influence of soil organic matter content on surveying and mapping features, which helps to accurately identify key factors affecting geological stability, soil fertility, etc., and adjust the semantic weights of surveying and mapping data accordingly, improving the applicability of surveying and mapping data. Traditional GIS systems often lack in-depth mining of complex geographical features when processing geological data. The present invention constructs a surveying and mapping enhanced semantic hypergraph, enabling surveying and mapping data to better express the features of plots in different environments and improving the scientific nature of land evaluation and management. In terms of data storage and management, the present invention uses adaptive sparse coding to optimize the information of plot types, making data storage more efficient and improving the real-time availability of surveying and mapping data. Compared with traditional paper archives or static GIS databases, the present invention can update the information of plot types in real time through the Internet of Things and provide more efficient data retrieval and analysis capabilities through the surveying and mapping enhanced semantic hypergraph. In addition, the storage method of real-time surveying and mapping stream data blocks optimizes the dynamic management ability of surveying and mapping data, enabling the construction land information management platform to continuously obtain the latest surveying and mapping data and provide more accurate decision-making support for applications such as environmental monitoring, overcoming the problem of decision-making delay caused by lagging data updates in traditional GIS systems.

[0011] Optionally, step S1 is specifically as follows:

[0012] Step S11: Obtain the construction land surveying and mapping data, and perform data preprocessing on the construction land surveying and mapping data to obtain the construction land surveying and mapping data to be analyzed;

[0013] Step S12: Perform modal decoupling and feature standardization on the construction land surveying and mapping data to be analyzed to obtain standardized surveying and mapping data;

[0014] Step S13: Conduct spatio-temporal feature correlation based on the standardized surveying and mapping data to obtain a surveying and mapping spatio-temporal feature map;

[0015] Step S14: Perform multi-modal feature fusion on the surveying and mapping spatio-temporal feature map and constrain the noise surveying and mapping bottleneck to obtain a preliminary perception fusion matrix;

[0016] Step S15: Calculate the information entropy distribution of the preliminary perception fusion matrix, and perform autoregressive completion and regional boundary optimization on the low-confidence data regions in the information entropy distribution to obtain a multi-modal perception matrix for construction land.

[0017] By obtaining the construction land surveying and mapping data and performing data preprocessing, the present invention can effectively clean and standardize the original data, remove measurement errors and noise, and ensure the accuracy of the data in subsequent analysis. The beneficial effect of this step is that it avoids the data inaccuracy caused by factors such as human errors and environmental interference in traditional surveying and mapping, thereby improving the overall data quality and laying a solid foundation for subsequent processing and analysis. By adopting the modal decoupling and feature standardization technology, different modal data (such as lidar, remote sensing images, etc.) in multi-source data can be effectively separated and standardized, enabling the data collected by different sensors to be compared and processed on a unified scale. This technology solves the problems of inconsistent data sources and large standardization differences in traditional surveying and mapping methods, and improves the consistency and comparability of the data. Further, through spatio-temporal feature correlation, a surveying and mapping spatio-temporal feature map is constructed, which can reveal the variation laws of the data in the time and space dimensions, providing in-depth insights for land planning, environmental monitoring, etc. Through the multi-modal feature fusion technology, by combining different types of data sources, the comprehensiveness and depth of the surveying and mapping data are enhanced. At the same time, constraining the noise surveying and mapping bottleneck effectively avoids information distortion and noise interference during the data fusion process, ensuring the high quality of the fused data. This process makes the subsequent data analysis and decision-making more reliable by improving the credibility of the data, providing more accurate information support for the management of construction land. By calculating the information entropy distribution of the preliminary perception fusion matrix and performing autoregressive completion and regional boundary optimization on the low-confidence data regions, this method demonstrates significant advantages in data repair and optimization. Especially in the case of missing or abnormal data, the autoregressive completion technology can accurately repair the low-confidence regions, eliminating potential errors caused by incomplete data, thereby enhancing the overall consistency and integrity of the data. In addition, the regional boundary optimization technology further improves the spatial resolution and accuracy of the surveying and mapping data, making the final multi-modal perception matrix for construction land more in line with the actual geographical features, greatly enhancing the applicability and reliability of the surveying and mapping data.

[0018] Optionally, step S14 is specifically as follows:

[0019] Step S141: Perform principal component dimensionality reduction based on the mapping spatio-temporal feature map, retain the feature components with a cumulative contribution rate ≥ 95%, and construct a mapping feature subspace;

[0020] Step S142: Perform adaptive weight attention multi-modal feature fusion according to the mapping feature subspace, dynamically adjust the feature contribution degrees of different mapping feature modalities in the mapping feature subspace, and obtain a multi-modal weighted fusion feature matrix;

[0021] Step S143: Detect low information gain regions based on the multi-modal weighted fusion feature matrix, and set the Softmax maximum probability < 0.6 to identify low confidence information regions in the multi-modal weighted fusion feature matrix;

[0022] Step S144: Combine the low confidence information regions and low information gain regions to perform high entropy anomaly detection, identify the mapping data noise bottleneck region, and perform adaptive multi-scale fusion to obtain a mapping bottleneck mask;

[0023] Step S145: Perform multi-modal feature bottleneck constraint on the multi-modal weighted fusion feature matrix according to the mapping bottleneck mask to obtain a preliminary perception fusion matrix.

[0024] In the present invention, the principal component dimensionality reduction technology reduces the data dimension, removes redundant information, streamlines the data structure, and at the same time retains the key features of the data by retaining the feature components with a cumulative contribution rate ≥ 95%, laying a more efficient foundation for subsequent analysis. Then, the adaptive weight attention multi-modal feature fusion can dynamically adjust the contribution degrees of different mapping modalities, making the role of each data type more accurate, thereby improving the effectiveness and accuracy of data fusion. Based on the setting of the low information gain region and the Softmax maximum probability threshold, the low confidence information regions in the mapping data can be effectively identified, and unreliable data can be effectively filtered out, thus avoiding the interference of low-quality data on the overall result. On this basis, through the high entropy anomaly detection technology, the noise bottleneck region is further identified and located, the data accuracy is improved, and the spatial resolution and detail performance of the data are enhanced through adaptive multi-scale fusion. Combining the mapping bottleneck mask for feature bottleneck constraint ensures that the key features in the data are fully retained, while the suppression of noise and redundancy parts guarantees the quality of the fused data.

[0025] Optionally, step S15 is specifically as follows:

[0026] Step S151: Calculate the information entropy distribution of the preliminary perception fusion matrix, set the entropy segmentation threshold [0.1, 0.5] to divide the high entropy region and the low entropy region, and obtain a mapping entropy partition map;

[0027] Step S152: Perform low-confidence time series interpolation on the low-entropy regions in the surveyed entropy partition map, set the confidence range as [0.5, 0.8], and set the spatial consistency coefficient as [0.1, 0.3] to optimize spatial consistency, obtaining a confidence-enhanced surveyed matrix;

[0028] Step S153: Perform local anomaly surveyed feature correction on the high-entropy regions in the surveyed entropy partition map, obtaining a noise-adaptive correction matrix;

[0029] Step S154: Combine the confidence-enhanced surveyed matrix and the noise-adaptive correction matrix for regional boundary regularization, obtaining a boundary-optimized surveyed matrix;

[0030] Step S155: Set a consistency evaluation coefficient [0.8, 1.0] to perform probability consistency evaluation on the boundary-optimized surveyed matrix, and perform adaptive confidence reallocation according to the probability consistency evaluation result, obtaining a multi-modal perception matrix for construction land.

[0031] Through the calculation of information entropy distribution and the division of entropy partitions, the present invention can effectively distinguish high-entropy and low-entropy regions, thereby identifying key differences and uncertain regions in the surveyed data. In high-entropy regions, the uncertainty of the data is relatively high. By using local anomaly surveyed feature correction, the interference of noise on data quality can be reduced, further improving the data accuracy. In low-entropy regions, through low-confidence time series interpolation and combined with spatial consistency optimization, missing data can be effectively filled, and the consistency and reliability of the data can be improved, especially in regions with significant spatial changes. Combining the confidence-enhanced surveyed matrix and the noise-adaptive correction matrix for regional boundary regularization effectively optimizes the regional boundary and reduces data errors caused by unclear boundaries. Finally, by performing probability consistency evaluation on the boundary-optimized surveyed matrix through the consistency evaluation coefficient, the confidence allocation can be automatically adjusted to ensure high consistency and high accuracy of the data. The combination of these steps improves the quality of the surveyed data, enabling the multi-modal perception matrix for construction land to more accurately reflect the real plot information.

[0032] Optionally, the multi-supervised deviation removal of the surveyed data described in step S2 is specifically:

[0033] Perform latent variable feature decomposition on the multi-modal perception matrix for construction land, and perform feature orthogonality to generate a surveyed latent variable feature set;

[0034] Perform unsupervised adversarial correction based on the surveyed latent variable feature set to construct a cross-device distribution adversarial network, and use the cross-device distribution adversarial network to perform maximum mean discrepancy alignment on the multi-modal perception matrix for construction land, obtaining aligned surveyed data;

[0035] Perform self-supervised abnormal area detection on the aligned surveying and mapping data, and set the abnormal area boundary in combination with the preset confidence threshold to obtain the surveying and mapping abnormal area mask;

[0036] The aligned surveying and mapping data are combined with the surveying and mapping anomaly area mask to compensate for the low-confidence data area errors and obtain the construction land perception tensor.

[0037] The present invention can effectively extract the core features in the multimodal perception matrix of construction land, remove redundant information, and improve the expression accuracy of data through latent variable feature decomposition and feature orthogonalization. This process helps to capture potential features that are valuable for surveying and mapping data analysis and enhance the effectiveness of data in subsequent analysis. The use of unsupervised adversarial correction and cross-device distributed adversarial network for maximum mean difference alignment helps to eliminate the distribution differences of surveying and mapping data between different devices, so that data from different devices are aligned under the same standard, improving the consistency and comparability of data. Self-supervised abnormal area detection can accurately identify abnormal areas in the data, and setting the abnormal area boundaries in combination with confidence thresholds can more accurately define the scope of the abnormal area and reduce the impact of errors. By performing error compensation on low-confidence data in abnormal areas, data missing can be effectively filled, the impact caused by data missing or noise can be reduced, and the integrity and accuracy of surveying and mapping data can be improved. The construction land perception tensor finally generated can more truly and accurately reflect the various characteristics of the construction land, ensuring the efficiency and accuracy of subsequent decision-making and analysis.

[0038] Optionally, the organic matter content causal reasoning described in step S4 is specifically:

[0039] Decompose the time series characteristics of the soil evolution trend of the construction land, extract the evolution characteristics of the soil organic matter content, and perform time window normalization to generate the soil evolution characteristic matrix;

[0040] The Granger causality test of the evolution characteristics is carried out according to the soil evolution characteristic matrix, and the preliminary causal relationship and preliminary causal weight of the soil organic matter content are set based on the causal test results, so as to generate a preliminary soil causal diagram;

[0041] The preliminary soil causal diagram was analyzed by disturbance variables, redundant factors were eliminated, and the causal effect matrix of soil organic matter content was calculated to construct a causal inference model;

[0042] The causal inference model is robustly corrected, and the weight distribution in the corrected causal effect matrix is ​​adjusted through the preset causal entropy weight strategy to obtain the optimized soil organic matter causal inference model.

[0043] Based on the optimized soil organic matter causal inference model, the standardized mapping tensor was reversed to quantify the influencing factors of mapping characteristics.

[0044] Through temporal feature decomposition and time window normalization, the present invention can extract the evolution characteristics of soil organic matter content, helping to accurately reflect the temporal characteristics of soil quality changes, ensuring the timeliness and consistency of data. This process helps to identify the long-term trends and periodic changes of soil organic matter, providing reliable basic data for subsequent analysis. Granger causality test can reveal the causal relationship between soil organic matter content and other variables, providing strong support for understanding the dynamics of soil evolution. Generating a preliminary soil quality causal diagram based on the results of the causality test helps to clarify the influence chain between various factors, forming a preliminary causal relationship framework. By removing redundant factors, perturbation variable analysis effectively reduces the complexity of the model, and by calculating the causal effect matrix, it accurately quantifies the causal effect of soil organic matter content, improving the accuracy and reliability of the model. The combination of robustness correction and causal entropy weight strategy effectively improves the stability of the model and its ability to handle abnormal data, and the optimized causal inference model can better reflect the change law of soil organic matter content. Finally, based on the optimized model, reverse deduction can be carried out to quantify the influencing factors of mapping features, providing a scientific basis for land resource management.

[0045] Optionally, adjusting the semantic weight of the standardized mapping tensor in step S4 specifically includes:

[0046] Constructing a preliminary semantic weight matrix based on the mapping feature influencing factors and the standardized mapping tensor;

[0047] Performing soil property correlation weighted analysis on the preliminary semantic weight matrix to identify the contribution degree of each mapping feature in the preliminary semantic weight matrix in reverse deduction, obtaining the mapping feature deduction contribution degree;

[0048] Adjusting the feature weights in the preliminary semantic weight matrix according to the mapping feature deduction contribution degree to obtain an optimized semantic weight matrix;

[0049] Performing gradient descent processing on the optimized semantic weight matrix to iteratively optimize the weight distribution, obtaining an optimized matrix of semantic weight distribution;

[0050] Performing a stability test on the optimized matrix of semantic weight distribution, and making an adaptive adjustment to the optimized matrix of semantic weight distribution according to the results of the stability test to obtain a semantic weight matrix;

[0051] Mapping the semantic weight matrix to the standardized mapping tensor to adjust the semantic weight of the mapping feature, obtaining a mapping enhanced semantic hypergraph.

[0052] By constructing a preliminary semantic weight matrix based on mapping feature impact factors and standardized mapping tensors, the present invention can assign preliminary semantic weights to each mapping feature, providing a clear basis for subsequent optimization. The soil property correlation weighted analysis helps to identify the contribution degrees of each mapping feature in the reverse deduction, so as to accurately evaluate the influence of different features on the results. This process can ensure the accuracy and rationality of the weight matrix, providing a scientific basis for the optimized semantic weights. Adjusting the feature weights in the preliminary semantic weight matrix helps to better reflect the actual influence of mapping data in different environments, ensuring the accurate description of soil properties and their changes. Further optimizing the weight distribution through gradient descent processing can efficiently improve the performance and prediction accuracy of the model, ensuring the convergence and accuracy of the weight adjustment process. The stability test and adaptive adjustment can enhance the robustness of the model, ensuring its stability under different environments or data changes, and preventing overfitting or instability. Finally, by mapping the optimized semantic weight matrix to the standardized mapping tensor, the semantic weights of mapping features can be finely adjusted to generate a mapping enhanced semantic hypergraph, improving the accuracy and reliability of land mapping data, and thus providing more accurate support for land management decisions.

[0053] Optionally, step S5 is specifically as follows:

[0054] Step S51: Perform mapping feature clustering based on the mapping enhanced semantic hypergraph to obtain mapping feature clustering data;

[0055] Step S52: Obtain the expert data of plot types, perform feature matching on the expert data of plot types and the mapping feature clustering data, and assign plot type labels according to the feature matching to obtain plot type label data;

[0056] Step S53: Perform sparse coding on the plot type label data to obtain sparse coding data;

[0057] Step S54: Obtain real-time construction land mapping data through the Internet of Things, and perform real-time plot type division and feature coding on the real-time construction land mapping data based on the sparse coding data to obtain real-time mapping stream data blocks;

[0058] Step S55: Upload the real-time mapping stream data blocks to the land information management platform for information storage tasks.

[0059] Through clustering mapping features based on a mapping-enhanced semantic hypergraph, the present invention can effectively perform structured processing on mapping data, accurately reveal the similarity between features, and thus help identify plot features with similar attributes. This process improves the operability and classification accuracy of the data, laying a foundation for the identification of plot types. By matching features with the data of plot type experts, appropriate labels can be assigned to each plot type, ensuring the accuracy of the labels and the credibility of the data. The generation of sparse-coded data helps compress redundant data, making the data more concise and efficient, and can effectively improve the computing and storage efficiency. Real-time acquisition of construction land mapping data and combining sparse coding for plot type division and feature coding enable the mapping process to reflect plot changes in real time, improve the timeliness and accuracy of mapping data, and ensure that the land information management platform receives the latest and most accurate data. Finally, uploading the real-time mapping stream data block to the platform for information storage can ensure the unified management and convenient access of the data, thereby improving the overall efficiency of the land information management system and promoting the timeliness and scientific nature of decision-making.

[0060] Optionally, this specification also provides an Internet-of-Things-based construction land mapping information management system for executing the Internet-of-Things-based construction land mapping information management method described above. The Internet-of-Things-based construction land mapping information management system includes:

[0061] A data acquisition module for obtaining construction land mapping data through the Internet of Things and performing collaborative perception fusion on the construction land mapping data to obtain a construction land multi-modal perception matrix;

[0062] A debiasing module for performing multi-supervised debiasing on the construction land multi-modal perception matrix to obtain a construction land perception tensor, and performing multi-source consistency correction on the construction land perception tensor to obtain a standardized mapping tensor;

[0063] A soil quality evolution trend analysis module for obtaining construction land soil quality data and analyzing the construction land soil quality evolution trend based on the construction land soil quality data;

[0064] A causal reasoning module for performing causal reasoning on the organic matter content based on the construction land soil quality evolution trend and the standardized mapping tensor to obtain a mapping feature influencing factor; adjusting the semantic weight of the standardized mapping tensor based on the mapping feature influencing factor to obtain a mapping-enhanced semantic hypergraph;

[0065] An information storage module for performing plot type adaptive sparse coding on the mapping-enhanced semantic hypergraph to obtain a real-time mapping stream data block and uploading it to the land information management platform to perform information storage tasks.

[0066] The Internet of Things-based construction land surveying and mapping information management system of the present invention can implement any one of the Internet of Things-based construction land surveying and mapping information management methods of the present invention. It is a medium for coordinating operations and signal transmission between various modules to complete the Internet of Things-based construction land surveying and mapping information management method. The internal modules of the system cooperate with each other, thereby improving the management efficiency and accuracy of surveying and mapping data management.

[0067] Optionally, this specification also provides a computer-readable storage medium with a computer program stored thereon. When the computer program is executed, it implements the Internet of Things-based construction land surveying and mapping information management method as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] By reading the detailed description of the non-limiting embodiments with reference to the following drawings, other features, objectives, and advantages of the present invention will become more apparent:

[0069] Figure 1 It is a schematic flowchart of the steps of the Internet of Things-based construction land surveying and mapping information management method of the present invention;

[0070] Figure 2 It is a detailed schematic flowchart of step S1 in the present invention;

[0071] Figure 3 It is a detailed schematic flowchart of step S5 in the present invention;

[0072] The implementation, functional characteristics, and advantages of the objectives of the present invention will be further described with reference to the embodiments and the drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0073] The technical method of the present invention will be clearly and completely described below with reference to the drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0074] In addition, the drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus their repeated description will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0075] It should be understood that although terms such as "first" and "second" may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly, the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0076] To achieve the above object, please refer to Figures 1 to 3 , the present invention provides a method for managing construction land surveying and mapping information based on the Internet of Things. The method includes the following steps:

[0077] Step S1: Obtain construction land surveying and mapping data through the Internet of Things, and perform collaborative perception fusion on the construction land surveying and mapping data to obtain a construction land multi-modal perception matrix;

[0078] In this embodiment, the surveying and mapping data of multiple sensors are obtained through Internet of Things technology. The sensors include ground sensors, lidar, data captured by drones, etc. The data generated by each sensor is transmitted to the central processing unit in real time through wireless communication. The data fusion process uses a weighted average model, and the form of the model is: fusion data matrix = ∑ i = 1 n w i ·X i ; where w i represents the weight of the i-th sensor, X i represents the surveying and mapping data of the i-th sensor, and the weight is determined by the accuracy of the sensor. The value range of w i is [0, 1], and ∑w i = 1. Finally, the obtained construction land multi-modal perception matrix can be represented as a multi-dimensional matrix M m×n , where m is the number of plots and n is the different sensor information dimensions of each plot.

[0079] Step S2: Perform multi-supervised debiasing on the construction land multi-modal perception matrix to obtain a construction land perception tensor, and perform multi-source consistency correction on the construction land perception tensor to obtain a standardized surveying and mapping tensor;

[0080] In this embodiment, debiasing processing and multi-source consistency correction are performed on the perception matrix. First, a supervised learning algorithm is used to perform debiasing processing on the construction land multi-modal perception matrix, and a random forest regression model (RF) is adopted. Assume that the output of the debiasing model is matrix M clean, each of its elements represents the perceived data after de - deviation processing. First, standardize the measurement values of each sensor, and then use the random forest algorithm to correct the sensor data to obtain the corrected data. Then, perform multi - source consistency correction, and this step uses the Kalman filter algorithm to optimize the spatio - temporal consistency. The formula of the Kalman filter algorithm is: where is the corrected state vector, K k is the Kalman gain, z k is the observation value, H k is the observation matrix. In this process, the construction land perception tensor is corrected, and a more consistent data matrix is obtained. The final output is a standardized mapping tensor, whose dimension is still m×n, but the deviation has been removed and the multi - source data consistency is ensured.

[0081] Step S3: Obtain the construction land soil quality data, and analyze the evolution trend of the construction land soil quality based on the construction land soil quality data;

[0082] In this embodiment, the construction land soil quality data is obtained through the Internet of Things, and data such as humidity, pH value, temperature, and organic matter content in the construction land soil quality data are used to construct a regression model. Use the LSTM model (long short - term memory neural network) for time - series analysis to predict the evolution trend of the soil. The form of the model is: y t = LSTM(X t-1 , X t-2 ,..., X t-p ); where y t is the predicted value at time point t, and X t-p represents the input data in the past p time steps. The input data for each time step includes soil humidity, pH, temperature, etc., and the output y t is the predicted soil change (such as the change trend of humidity). These predicted data can be used to analyze the evolution trend of the soil, and generate a soil quality evolution trend matrix M soil trend , whose form is an m×p matrix, where m is the number of plots and p is the predicted time step length.

[0083] Step S4: Conduct causal inference on the organic matter content based on the construction land soil quality evolution trend and the standardized mapping tensor to obtain the geological organic matter content influencing factor; adjust the semantic weight of the standardized mapping tensor based on the geological organic matter content influencing factor to obtain a mapping enhanced semantic hypergraph;

[0084] In this embodiment, combining the soil quality evolution trend and the standardized mapping tensor, analyze the influence of the organic matter content through a causal inference model. Use the Granger causality test and the Bayesian network model to explore the causal relationship between soil characteristics and organic matter content. The specific formula of the causal inference is: y organic= f(X soil , X temperature , X humidity ); where y organic is the organic matter content, and X soil , X temperature , X humidity represent soil characteristics such as soil type, temperature, and humidity, respectively. Use the causal inference result to generate a geological organic matter content influence factor matrix. The dimension of this matrix is n×1, where n is the number of soil characteristics, and each element represents the influence factor of the corresponding characteristic on the organic matter content. Subsequently, the standardized mapping tensor is weighted and adjusted according to this influence factor. The specific semantic weight adjustment formula is: M adjusted = M standardized ·diag(w); where w is the geological organic matter content influence factor, and diag(w) is to convert the influence factor into a diagonal matrix for weighting. Finally, the mapping enhanced semantic hypergraph G enhanced is obtained. This hypergraph represents the complex relationships between plots and contains the correlations between soil characteristics and organic matter content.

[0085] Step S5: Perform plot type adaptive sparse coding on the mapping enhanced semantic hypergraph to obtain real-time mapping stream data blocks, and upload them to the land information management platform for information storage tasks.

[0086] In this embodiment, perform plot type adaptive sparse coding on the mapping enhanced semantic hypergraph and use the K-SVD algorithm for sparse coding. Assume the formula for the coding process is: X encoded = Sparse(G enhanced , λ); where λ is the sparsity parameter that controls the data compression rate after coding. The data block X encoded after sparse coding is further divided into real-time mapping stream data blocks and protected by AES encryption to ensure data security. Finally, the mapping data blocks are uploaded to the land information management platform for storage, forming a real-time land data storage system. All mapping data is managed and queried through the information storage platform to ensure the efficient storage and accurate use of land information.

[0087] Optionally, step S1 is specifically:

[0088] Step S11: Obtain construction land mapping data and perform data preprocessing on the construction land mapping data to obtain the construction land mapping data to be analyzed;

[0089] In this embodiment, the construction land surveying and mapping data of multiple sensors is obtained through the Internet of Things (IoT) platform. These sensors include terrestrial lidar, image data captured by drones, temperature and humidity data obtained by ground sensors, etc. The sensor data is transmitted to the central data processing unit in real time through a wireless network, and the data format is a two-dimensional array D raw (with dimensions of m×n, where m is the number of plots and n is the different feature dimensions of each plot). For data preprocessing, first, missing value processing and filtering are performed on the original data, and a simple mean imputation method and Gaussian filtering are used to remove noise. Then, unified timestamp alignment is performed for each data source to ensure the temporal consistency of the data, thereby obtaining the construction land surveying and mapping data to be analyzed. The preprocessed data matrix D preprocessed still has dimensions of m×n, where m is the number of plots and n is the set of all features.

[0090] Step S12: Perform modal decoupling and feature standardization on the construction land surveying and mapping data to be analyzed to obtain standardized surveying and mapping data;

[0091] In this embodiment, for the construction land surveying and mapping data D preprocessed modal decoupling is performed. The principal component analysis (PCA) method is used to reduce the dimension of the data and extract the most representative features in each mode. In this process, the obtained modal matrix is D modes , where D modes ={D1, D2,..., D p}, and each D i corresponds to a data mode with dimensions of m×n i , where n i is the number of features in this mode. Then, Z-score standardization processing is performed on each mode so that the mean of each mode is 0 and the standard deviation is 1. The standardized data matrix is D standardized , with dimensions of m×n, and the distribution of all features has been adjusted to a unified standard for subsequent processing.

[0092] Step S13: Perform spatio-temporal feature association based on the standardized surveying and mapping data to obtain a surveying and mapping spatio-temporal feature map;

[0093] In this embodiment, based on D standardized , spatio-temporal feature association is performed. For this purpose, the spatio-temporal graph convolutional network (ST-GCN) method is used, where each plot is regarded as a graph node, and the connections between nodes represent the relationship of adjacent plots in space. To represent temporal information, we add timestamp information to the feature vector of each node to form the spatio-temporal feature map G spatial-temporal , and the adjacency matrix A of this graph spatial-temporalDescribes the spatial dependence relationships among nodes and captures the temporal dependencies through convolution operations. Through the learning of the spatio-temporal convolutional network, a spatio-temporal feature map is obtained, which has m nodes and an adjacency matrix of m×m. Each node in the map contains the combined spatial and temporal features.

[0094] Step S14: Perform multi-modal feature fusion on the surveyed spatio-temporal feature map and constrain the noise survey bottleneck to obtain a preliminary perception fusion matrix.

[0095] In this embodiment, perform multi-modal feature fusion on the surveyed spatio-temporal feature map G spatial-temporal Specifically, the weighted fusion of each modal feature is performed through the self-attention mechanism. In this process, the multi-head attention mechanism is used to perform weighted combination on different modal features. Through the learned weight w i and the data D of each modality i , the fused matrix M fusion is obtained. The dimension of this matrix is m×n, where m is the number of plots and n is the feature dimension of each plot. During the fusion process, noise suppression constraints are also used to reduce the influence of noise on the final result through the L2 regularization and Dropout methods. Further, bottleneck constraints are used to suppress the noise in the feature map to ensure that the finally obtained perception fusion matrix is more accurate.

[0096] Step S15: Calculate the information entropy distribution of the preliminary perception fusion matrix and perform autoregressive completion and regional boundary optimization on the low-confidence data regions in the information entropy distribution to obtain a multi-modal perception matrix for construction land.

[0097] In this embodiment, calculate the information entropy distribution of the preliminary perception fusion matrix M fusion and perform autoregressive completion on the low-confidence regions in the information entropy distribution. First, calculate the information entropy of each data point in the matrix. The formula is: where p(x i ) represents the probability distribution of the data point x i , and H(x) is the entropy value of the data point. Through the calculation of the information entropy, the matrix H entropy is obtained, which reflects the confidence of the data of each plot. Then, perform autoregressive completion on the low-confidence regions. In this step, the ARIMA (Autoregressive Integrated Moving Average) model is used to predict based on historical data to fill in the missing values in the low-confidence regions. Specifically, the time series data is fitted by the autoregressive model and the values of the missing regions are predicted, and finally the completed matrix M complete completes the repair and optimization of all low-confidence data, which is suitable for further data analysis and decision support.

[0098] Optionally, step S14 is specifically as follows:

[0099] Step S141: Perform principal component dimensionality reduction based on the mapping spatio-temporal feature map, retain the feature components with a cumulative contribution rate ≥ 95%, and construct a mapping feature subspace;

[0100] In this embodiment, principal component analysis (PCA) dimensionality reduction is performed on the mapping spatio-temporal feature map to extract the most representative features and construct a mapping feature subspace. Assume that the mapping spatio-temporal feature map G spatial-temporal contains spatio-temporal features of multiple modalities, and the dimension of each modality is n i (where i = 1, 2,..., p, and p is the number of modalities). These features are dimensionally reduced by PCA, and the goal is to retain the feature components with a total cumulative contribution rate ≥ 95%. Specifically, the feature matrix G spatial-temporal is subjected to singular value decomposition (SVD): G spatial-temporal = U∑V T ; where U is the feature vector matrix, ∑ is the singular value matrix, and V T is the eigenvalue matrix. By calculating the contribution rate of each principal component, the feature components with a cumulative contribution rate reaching more than 95% are selected to generate the dimensionally reduced feature matrix G reduced , whose dimension is m×k, where k is the number of features retained after dimensionality reduction. The finally obtained mapping feature subspace F subspace contains the most representative spatio-temporal features and can be used for subsequent feature fusion operations.

[0101] Step S142: Perform adaptive weight attention multi-modal feature fusion according to the mapping feature subspace, dynamically adjust the feature contribution degrees of different mapping feature modalities in the mapping feature subspace, and obtain a multi-modal weighted fusion feature matrix;

[0102] In this embodiment, adaptive weight attention multi-modal feature fusion is performed according to the mapping feature subspace F subspace . For this purpose, the self-attention mechanism is used to perform weighted fusion on the features of each modality. Specifically, the self-attention mechanism is first applied to the feature vectors of each modality to calculate the weight α i of each modality. This weight value represents the contribution degree of the features of this modality to the final fusion result. Specifically, the multi-head self-attention mechanism is used to calculate the correlation between the features F i of each modality and the features of other modalities: where Q is the query vector, K is the key vector, V is the value vector, and d k is the dimension of the key vector. By calculating the attention weights of each modality, the weighted fusion matrix F fusion, with dimensions of m×n, where m is the plot and n is the number of fused features. Finally, the obtained multi-modal weighted fusion feature matrix F fusion will dynamically adjust the contribution degrees of different modalities, thereby improving the accuracy of feature fusion.

[0103] Step S143: Detect low-information-gain regions based on the multi-modal weighted fusion feature matrix, and set the Softmax maximum probability <0.6 to identify low-confidence information regions in the multi-modal weighted fusion feature matrix;

[0104] In this embodiment, based on the multi-modal weighted fusion feature matrix F fusion detect low-information-gain regions. To achieve this goal, first calculate the information gain of each feature, and its formula is: I(X) = H(X) - H(X|Y); where, I(X) is the information gain, H(X) is the entropy of feature X, and H(X|Y) is the conditional entropy after given feature Y. According to the calculation results, identify the feature regions with low information gain. To further identify low-confidence regions, use the Softmax function to calculate the maximum probability P max of each feature, and set the region with the maximum probability less than 0.6 as the low-confidence region: P max = max(Softmax(F fusion )); if P max <0.6, then this region is marked as a low-confidence region. Finally, the low-information-gain regions and low-confidence regions together constitute the detected suspected low-information-gain regions.

[0105] Step S144: Combine the low-confidence information regions and low-information-gain regions to perform high-entropy anomaly detection, identify the noise bottleneck regions in the surveying and mapping data, and perform adaptive multi-scale fusion to obtain the surveying and mapping bottleneck mask;

[0106] In this embodiment, combine the low-confidence information regions and low-information-gain regions to perform high-entropy anomaly detection to identify the noise bottleneck regions in the surveying and mapping data. Specifically, first calculate the entropy values of these low-confidence regions and low-information-gain regions, and calculate the entropy of each region through the formula: High-entropy regions usually indicate anomalies or noises, so focus on these regions. Then, apply the multi-scale fusion method, combine image processing techniques at different scales, and repair the noise regions through a multi-scale convolutional network. Specifically, use different convolutional kernel sizes to capture noise features at different scales, thereby performing efficient noise repair and data optimization. Finally, obtain the surveying and mapping bottleneck mask M mask , which marks all noise bottleneck regions and helps with subsequent data repair.

[0107] Step S145: Perform multi-modal feature bottleneck constraint on the multi-modal weighted fusion feature matrix according to the mapping bottleneck mask to obtain a preliminary perception fusion matrix.

[0108] In this embodiment, according to the mapping bottleneck mask M mask perform multi-modal feature bottleneck constraint on the multi-modal weighted fusion feature matrix F fusion . To reduce the influence of noise and low-confidence data, the low-confidence region is constrained by a weighted penalty method. The specific operation is: F final = F fusion ×(1 - M mask ); where M mask is the mapping bottleneck mask. If a certain region is a noise region, the corresponding value is 1, otherwise it is 0. Through this constraint, the influence of noise on the data can be effectively eliminated, and a more accurate perception fusion result can be obtained. Finally, the preliminary perception fusion matrix F final after bottleneck constraint can be used for subsequent analysis and decision support.

[0109] Optionally, step S15 is specifically:

[0110] Step S151: Calculate the information entropy distribution of the preliminary perception fusion matrix, set the entropy division threshold [0.1, 0.5] to divide the high-entropy region and the low-entropy region, and obtain the mapping entropy division map;

[0111] In this embodiment, calculate the information entropy distribution of the preliminary perception fusion matrix F final to divide the high-entropy region and the low-entropy region of the mapping data. The information entropy H(X) reflects the uncertainty of the mapping data, and its calculation formula is as follows: where p(x i ) represents the probability distribution of the mapping feature value x i appearing. Calculate the local information entropy of each region of F final , and set the entropy division threshold range as [0.1, 0.5] to distinguish the high-entropy region and the low-entropy region. For the region where the information entropy value H(X) is lower than 0.1, it is marked as the low-entropy region, while the region higher than 0.5 is marked as the high-entropy region, and the remaining regions are not adjusted. Finally, based on the entropy value distribution of the mapping data, construct the mapping entropy division map M entropy , where each grid cell is labeled as a high-entropy (greater influence of noise) or low-entropy (higher data consistency) category for subsequent adaptive correction.

[0112] Step S152: Perform low-confidence time series interpolation on the low-entropy region in the mapping entropy division map, set the confidence range as [0.5, 0.8], and set the spatial consistency coefficient as [0.1, 0.3] to optimize the spatial consistency, and obtain the confidence-enhanced mapping matrix;

[0113] In this embodiment, for the mapping entropy partition map M entropy in the low entropy region, low-confidence time series interpolation is performed to optimize the spatial consistency of this region. The low entropy region represents better data quality but may have local missing values, so a time series interpolation method is used for repair. Assuming that the confidence distribution of the low entropy region is C(X), the confidence range is set to [0.5, 0.8], and the mapping data within this range is interpolated. The specific interpolation method uses Bayesian Optimal Interpolation (BOI), and its mathematical form is as follows: X interp =X obs +K(X prior -X obs ); where X obs is the observed data, X prior is the prior estimate value, K is the Kalman gain coefficient, which adjusts the credibility of the interpolated data. In addition, the range of the spatial consistency coefficient is set to [0.1, 0.3], and the spatial Laplacian regularization method is used to optimize the interpolation result to enhance the consistency of the mapping data, and finally a confidence-enhanced mapping matrix M conf is generated.

[0114] Step S153: Perform local abnormal mapping feature correction on the high entropy region in the mapping entropy partition map to obtain a noise adaptive correction matrix;

[0115] In this embodiment, local abnormal mapping feature correction is performed on the high entropy region in the mapping entropy partition map M entropy . Since there may be noise, abnormal mapping values or equipment errors in the high entropy region, an adaptive filtering strategy needs to be used for data correction. The specific method is to construct an Adaptive Mean Filter (AMF), and its update formula is as follows: where X i is the mapping value of the high entropy region, w i is the weight factor related to the confidence, and N is the neighborhood window size (set to 3×3 or 5×5). When the entropy value of a certain region is high and the mapping value deviates from the neighborhood mean by more than 2 standard deviations, it is considered that there is an abnormality in this region, and AMF is used for smoothing correction. In addition, to avoid edge effects, a gradient constraint mechanism is adopted during data update, so that the adjusted data will not affect the mapping boundary. Finally, the corrected mapping data forms a noise adaptive correction matrix M denoise .

[0116] Step S154: Combine the confidence-enhanced mapping matrix and the noise adaptive correction matrix for regional boundary regularization to obtain a boundary optimized mapping matrix;

[0117] In this embodiment, the confidence-enhanced mapping matrix M conf and the noise-adaptive correction matrix M denoise are combined to perform regional boundary regularization to optimize the boundary continuity of the mapping data. Since the confidence-enhanced mapping matrix mainly optimizes low-entropy regions, while the noise-adaptive correction matrix targets high-entropy regions, it is necessary to fuse the two to ensure the smoothness of data transition. The specific method uses second-order total variation regularization (STV), and its optimization objective is: where X is the mapping data after boundary optimization. The goal of constraint optimization is to reduce data mutations and make the boundary smoother. Subsequently, a boundary gradient constraint is further introduced to ensure that the boundary region will not be over-smoothed: where G boundary is the boundary gradient change. By adjusting the gradient threshold of the boundary region, the integrity of the feature contour can be maintained while eliminating boundary anomalies in the mapping data. Finally, after regularization processing, the boundary-optimized mapping matrix M boundary is obtained.

[0118] Step S155: Set a consistency evaluation coefficient [0.8, 1.0] to perform probability consistency evaluation on the boundary-optimized mapping matrix, and perform adaptive confidence reassignment according to the probability consistency evaluation result to obtain a multi-modal perception matrix for construction land.

[0119] In this embodiment, probability consistency evaluation is performed on the boundary-optimized mapping matrix M boundary , and the confidence assignment is adaptively adjusted according to the evaluation result. Finally, a multi-modal perception matrix M final for construction land is obtained. Set the consistency evaluation coefficient range to [0.8, 1.0] to measure the consistency of the mapping data. The evaluation method is based on the Mahalanobis distance (MD), and its calculation formula is: where μ is the mean of the mapping data, ∑ is the covariance matrix, and X is the data to be evaluated. If the Mahalanobis distance of a certain mapping area is lower than the threshold 0.8, it means that the consistency of this area is high, and the confidence of this area is increased; if the Mahalanobis distance is higher than 1.0, it indicates that there are anomalies in the data of this area, and its confidence is reduced. To optimize the final confidence assignment, the Dirichlet distribution is used for adaptive confidence adjustment, and its calculation formula is: where C i is the confidence of the mapping data, and α i is the adjustment parameter. Finally, after probability consistency evaluation and adaptive adjustment, a multi-modal perception matrix M final, the matrix is optimized in terms of information consistency, boundary integrity, and mapping confidence, and can be used for subsequent mapping analysis and planning applications.

[0120] Optionally, the multi-supervised debiasing of the mapping data described in step S2 is specifically as follows:

[0121] Perform latent variable feature decomposition on the multi-modal perception matrix of construction land and perform feature orthonormalization to generate a mapping latent variable feature set;

[0122] In this embodiment, perform non-negative matrix factorization (NMF) on the multi-modal perception matrix M of construction land final and set the decomposition rank k = 10 to extract a low-dimensional latent variable feature matrix: where W represents the latent variable representation of different mapping modes, and H represents the contribution degree of the latent variable in different regions. Subsequently, perform Gram-Schmidt process orthonormalization on W: W′ = GS(W) to ensure the orthogonality between latent variable features and avoid feature redundancy. Finally, generate the mapping latent variable feature set M Iv .

[0123] Perform unsupervised adversarial correction based on the mapping latent variable feature set to construct a cross-device distribution adversarial network, and use the cross-device distribution adversarial network to perform maximum mean discrepancy alignment on the multi-modal perception matrix of construction land to obtain aligned mapping data;

[0124] In this embodiment, aiming at the differences in data distribution among different mapping devices (such as lidar, satellite remote sensing, and UAV mapping), construct a cross-device adversarial network, including a generator G(x) and a discriminator D(x). Among them, the generator is used to generate mapping data with a unified distribution, and the discriminator is used to distinguish the data distributions from different devices. The network loss function is optimized using the maximum mean discrepancy (MMD): where φ(x) represents the Gaussian kernel mapping function, and N and M are the numbers of data samples from different devices respectively. During the training process, optimize the generator through backpropagation to make the distribution generated by G(x) as close as possible to the distribution of the real mapping data, and finally generate the aligned mapping data M A .

[0125] Perform self-supervised abnormal region detection on the aligned mapping data, and set the boundary of the abnormal region in combination with a preset confidence threshold to obtain a mapping abnormal region mask;

[0126] In this embodiment, perform abnormal region detection based on the self-supervised Transformer structure. First, perform local feature extraction using a sliding window size w = 5: F local = CNN(M A , w); then, input the local features into the Transformer network to calculate the abnormal score S(mi，j ):S(m i，j ) = Transformer(F local ); Set the abnormal confidence threshold to 0.9. If S(m i，j ) > 0.9, then mark this point as an abnormal point. The set of all abnormal points constitutes the mapping anomaly region mask M M .

[0127] Combine the mapping anomaly region mask to perform error compensation on the aligned mapping data for the low-confidence data region, and obtain the construction land perception tensor.

[0128] In this embodiment, error compensation is performed on the low-confidence region (i.e., the abnormal region) in M A . An error compensation strategy based on Krigin g interpolation is adopted to estimate the values in the low-confidence region: ∑ k ω k = 1; where ω k is calculated from the Kriging spatial correlation. The compensated data matrix is defined as: The obtained construction land perception tensor not only retains the accuracy of the high-confidence data but also corrects the errors in the abnormal region.

[0129] Optionally, the causal inference of the organic matter content described in step S4 is specifically:

[0130] Perform time series feature decomposition on the evolution trend of the construction land soil quality, extract the evolution characteristics of the soil organic matter content, and perform time window normalization to generate the soil quality evolution feature matrix;

[0131] In this embodiment, for key parameters such as soil organic carbon (SOC), total nitrogen (TN), and soil water content (SWC) included in the evolution trend of the construction land soil quality, a time series data set is constructed: S T = {s1, s2,..., s T}; where s i represents the soil sample data at the i-th time step. Empirical mode decomposition (EMD) is used to perform time series feature decomposition on S T to extract the evolution trends at different scales: where IMF j represents the j-th intrinsic mode function, and R is the residual term. Then, the main k = 5 feature components are selected for principal component analysis (PCA) dimensionality reduction to obtain the evolution characteristics of the soil organic matter content Next, calculate the local normalization value within a sliding time window of 12 months: The normalized data constitutes the soil quality evolution feature matrix M Q .

[0132] Perform the Granger causality test on the evolution characteristics based on the soil quality evolution characteristic matrix, and set the preliminary causal relationship and preliminary causal weight of the soil organic matter content based on the results of the causality test, so as to generate a preliminary soil quality causal diagram;

[0133] In this embodiment, the Granger causality test is used to evaluate the causal relationship between soil variables. Set the maximum lag order p = 3, and perform two-way tests on each feature pair (X, Y) in M Q Calculate the causal strength through the F statistic. If the p-value is lower than the significance level (α = 0.05), it is considered that there is a Granger causal relationship between variable X and Y, and the causal weight ω(X, Y) is recorded. Construct a preliminary soil quality causal diagram G C =(V, E, W), where V represents the set of soil variable nodes, E is the set of causal relationship edges, and W is the causal weight matrix.

[0134] Perform perturbation variable analysis on the preliminary soil quality causal diagram, eliminate redundant factors, and calculate the causal effect matrix of the soil organic matter content, so as to construct a causal inference model;

[0135] In this embodiment, perform perturbation variable analysis on the causal diagram G C First, calculate the SHAP value (Shapley Additive Explanations) of each variable using random forest regression, screen out redundant variables with an importance threshold lower than 0.02, and delete their corresponding edges E r : G′ C =(V - V r , E - E r , W - W r ); where V r is the set of redundant variables, and W r is the weight to be eliminated. Then, calculate the causal effect matrix E c of the soil organic matter content based on structural equation modeling (SEM), and estimate its coefficients using the least squares method: E C =(I - B) -1 Γ; where B is the regression coefficient matrix between variables, and Γ is the influence weight of the observation error. Construct a causal inference model M c , let the mapping variable set be X′ = {x1, x2,..., x n}, the soil organic matter content S O , and map the variable S O = f(X′, E C ) + ∈; where f(·) is a non-linear regression function, is a Gaussian noise term. The causal model parameters are optimized using maximum likelihood estimation (MLE), and 5-fold cross-validation is performed to improve the generalization ability of the model. Finally, the constructed causal inference model M c can be used to predict the evolution trend of soil organic matter content and provide theoretical support for subsequent steps (such as causal effect optimization and reverse deduction of mapping features).

[0136] The causal inference model is robustly corrected, and the weight distribution in the corrected causal effect matrix is adjusted through a preset causal entropy weight strategy to obtain an optimized soil organic matter causal inference model;

[0137] In this embodiment, to enhance the robustness of the causal inference model, Poisson Regression is used to correct the bias of the causal effect matrix E C : where λ is the Poisson distribution parameter and β is the correction coefficient. Subsequently, the causal entropy weight strategy is used to reallocate the causal weights. Calculate the causal entropy of each variable: H(X i ) = -∑ j p ij logp ij ; then, the entropy weights are assigned to the causal matrix: Finally, the optimized causal inference model M C can more accurately predict the change trend of soil organic matter content.

[0138] Based on the optimized soil organic matter causal inference model, reverse deduction is performed on the standardized mapping tensor to quantify the mapping feature influence factors.

[0139] In this embodiment, in the optimized causal inference model M C reverse deduction is performed on the standardized mapping tensor , where m is the number of grid cells in the mapping area and n is the mapping feature dimension (such as terrain slope, elevation, vegetation coverage, etc.). First, based on the optimized causal effect matrix E″ C calculate the influence of soil organic matter content S O on the mapping feature T S through the reverse causal deduction equation: where S O represents the standardized data of soil organic matter content, is the transpose of the optimized causal matrix, indicating the causal influence of soil organic matter content on the mapping features. Then, calculate the influence factor matrix I F : where I F represents the relative influence weight of the normalized soil organic matter content on each mapping feature. Finally, the influence factor matrix I TIt can be used to quantify the functional relationship between changes in soil organic matter content and mapping parameters such as topographic features, vegetation indices, and surface temperature, providing accurate data support for soil quality assessment and land use planning.

[0140] Optionally, adjusting the semantic weight of the standardized mapping tensor described in step S4 specifically includes:

[0141] Constructing a preliminary semantic weight matrix based on the mapping feature influence factor and the standardized mapping tensor;

[0142] In this embodiment, let the standardized mapping tensor be where m is the number of grid cells in the mapping area, and n is the dimension of mapping features (such as topographic slope, elevation, vegetation coverage, etc.). The mapping feature influence factor I F quantifies the influence intensity of each mapping feature on the soil organic matter content. Through normalization, a preliminary semantic weight matrix W O is constructed, and its elements are defined as: where w ij represents the semantic similarity weight between mapping feature i and mapping feature j, and the weight range is set between [0, 1].

[0143] Performing soil property correlation weighted analysis on the preliminary semantic weight matrix to identify the contribution degree of each mapping feature in the reverse deduction in the preliminary semantic weight matrix, and obtaining the mapping feature deduction contribution degree;

[0144] In this embodiment, Pearson correlation analysis is used to calculate the correlation between the preliminary semantic weight matrix W O and soil property variables (such as soil moisture, density), and a mapping feature deduction contribution degree vector C p is defined as: where, represents the correlation coefficient between mapping feature i and the soil organic matter content. If the contribution degree is higher than the threshold of 0.6, it is considered that this feature plays an important role in the reverse deduction.

[0145] Adjusting the feature weights in the preliminary semantic weight matrix according to the mapping feature deduction contribution degree to obtain an optimized semantic weight matrix;

[0146] In this embodiment, according to the normalized value of C D , the weights of W O are adjusted: W1 = W0 ⊙ C P ; where ⊙ is element-wise multiplication. In this way, the semantic weights of high-contribution mapping features can be enhanced, and the influence of low-contribution features can be weakened, obtaining an optimized semantic weight matrix W1.

[0147] Performing gradient descent processing on the optimized semantic weight matrix to iteratively optimize the weight distribution and obtain a semantic weight distribution optimization matrix;

[0148] In this embodiment, a loss function is defined: Among them, is the updated weight, and λ is the regularization coefficient (such as 0.01). The gradient descent iteration is performed using the Adam optimizer: where α is the learning rate (such as 0.05). Finally, a converged optimized matrix W2 of the semantic weight distribution is obtained.

[0149] Perform a stability test on the optimized matrix of the semantic weight distribution, and adaptively adjust the optimized matrix of the semantic weight distribution according to the stability test result to obtain a semantic weight matrix;

[0150] In this embodiment, the Markov stability analysis method is used to calculate the stability score of the matrix W2 at different time steps: If S t < 0.01, the matrix converges; otherwise, adjust the regularization parameter λ, return to the previous step for further optimization, and finally obtain the semantic weight matrix W3.

[0151] Map the semantic weight matrix to the standardized mapping tensor, and adjust the semantic weights of the mapping features to obtain a mapped enhanced semantic hypergraph.

[0152] In this embodiment, the standardized mapping tensor T S is weighted using the semantic weight matrix W3: T' M = T M ⊙W3; further construct a hypergraph H, where the hyperedge e i is aggregated by mapping features with high semantic similarity, and the hypergraph Laplacian matrix is defined: where D H is the hyperedge degree matrix. Finally, the mapped enhanced semantic hypergraph H is obtained.

[0153] Optionally, step S5 is specifically as follows:

[0154] Step S51: Perform mapping feature clustering based on the mapped enhanced semantic hypergraph to obtain mapping feature clustering data;

[0155] In this embodiment, the mapped enhanced semantic hypergraph is used to cluster the features of the construction land mapping data. First, based on the weight relationship in the semantic hypergraph, the mapping features are classified according to similarity. The K-Means clustering method is used to group the mapping features in the hypergraph to determine the mapping feature categories of different plots. During the clustering process, the number of clustering clusters k = 5 is set, and the belonging category of each feature is iteratively calculated. Finally, a mapping feature clustering data set is formed, and each clustering result represents a set of mapping features of a type of plot.

[0156] Step S52: Obtain the expert data of plot types, perform feature matching on the expert data of plot types and the clustering data of surveying and mapping features, and assign plot type labels based on the feature matching to obtain plot type label data;

[0157] In this embodiment, obtain the labeled plot type data from the land and resources database or the manual surveying and mapping expert system, including the name of the plot, the use (such as cultivated land, construction land, forest land), and the surveying and mapping features. By calculating the similarity between the clustering data of surveying and mapping features and the expert data, screen the most matching plot type, and assign the corresponding plot type label to the clustered surveying and mapping data. For example, if the plot type corresponding to a certain type of surveying and mapping feature is mainly "construction land", then all data of this type are given the label of "construction land" to form a plot type label data set.

[0158] Step S53: Perform sparse coding on the plot type label data to obtain sparse coding data;

[0159] In this embodiment, the plot type label data is processed by sparse coding. Through the One-Hot Encoding method, each plot type is converted into a binary array. For example, if the plot types include "cultivated land", "construction land", and "forest land", then "construction land" can be encoded as [0, 1, 0]. This can ensure that the data can be effectively matched and calculated in subsequent processing, while reducing the consumption of computing resources and improving the data retrieval efficiency.

[0160] Step S54: Obtain the real-time construction land surveying and mapping data through the Internet of Things, and perform real-time plot type classification and feature coding on the real-time construction land surveying and mapping data based on the sparse coding data to obtain real-time surveying and mapping data blocks;

[0161] In this embodiment, use Internet of Things technologies (such as drone surveying and mapping, satellite remote sensing, GNSS positioning, etc.) to obtain the real-time construction land surveying and mapping data. According to the sparse coding data, automatically classify the real-time obtained surveying and mapping data to determine the plot type it belongs to. During the classification process, use the similarity calculation method to match the surveying and mapping data to the most similar category, and perform feature coding to ensure that the real-time data stream can be quickly associated with the corresponding plot category, and finally generate real-time surveying and mapping data blocks. For example, if the height and terrain features of a certain surveying and mapping data point are most similar to "construction land", it is automatically marked as "construction land" and added to the real-time data stream.

[0162] Step S55: Upload the real-time surveying and mapping data blocks to the land information management platform to perform information storage tasks.

[0163] In this embodiment, after the real-time mapping stream data blocks are sorted, they are uploaded to the land information management platform through a data interface. The platform supports structured storage, classifies and stores information according to plot numbers, timestamps, plot types, etc., and provides subsequent query and analysis capabilities. For example, the real-time mapping data of a certain plot is uploaded every 10 seconds to ensure timely data updates and provide data support for land planning, management, and decision-making.

[0164] Optionally, this specification also provides an Internet of Things-based construction land mapping information management system for implementing the Internet of Things-based construction land mapping information management method described above. The Internet of Things-based construction land mapping information management system includes:

[0165] A data acquisition module for obtaining construction land mapping data through the Internet of Things and performing collaborative perception fusion on the construction land mapping data to obtain a construction land multi-modal perception matrix;

[0166] A de-biasing module for performing multi-supervised de-biasing on the construction land multi-modal perception matrix for mapping data to obtain a construction land perception tensor, and performing multi-source consistency correction on the construction land perception tensor to obtain a standardized mapping tensor;

[0167] A soil quality evolution trend analysis module for obtaining construction land soil quality data and analyzing the construction land soil quality evolution trend based on the construction land soil quality data;

[0168] A causal reasoning module for performing causal reasoning on the organic matter content according to the construction land soil quality evolution trend and the standardized mapping tensor to obtain a mapping feature influence factor; adjusting the semantic weight of the standardized mapping tensor based on the mapping feature influence factor to obtain a mapping enhanced semantic hypergraph;

[0169] An information storage module for performing plot type adaptive sparse coding on the mapping enhanced semantic hypergraph to obtain real-time mapping stream data blocks and uploading them to the land information management platform to perform information storage tasks.

[0170] Optionally, this specification also provides a computer-readable storage medium with a computer program stored thereon. When the computer program is executed, it implements the Internet of Things-based construction land mapping information management method described above.

[0171] Therefore, from any perspective, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the application documents are intended to be encompassed within the present invention.

[0172] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features invented herein.

Claims

1. A construction land surveying and mapping information management method based on the Internet of Things, characterized in that: The following steps are involved: Step S1: Acquire construction land surveying and mapping data through the Internet of Things, and perform collaborative perception fusion on the construction land surveying and mapping data to obtain a multi-modal perception matrix of the construction land; Step S2: Perform multi-supervision debiasing on the multimodal perception matrix of construction land to obtain a perception tensor of construction land, and perform multi-source consistency correction on the perception tensor of construction land to obtain a standardized mapping tensor; Step S3: Acquire soil quality data of construction land, and analyze the soil quality evolution trend of construction land based on the soil quality data of construction land; Step S4: performing causal reasoning on organic matter content according to the soil evolution trend of the construction land and the standardized surveying and mapping tensor to obtain the influencing factor of geological organic matter content; adjusting the semantic weight of the standardized surveying and mapping tensor based on the influencing factor of geological organic matter content to obtain a surveying and mapping enhanced semantic hypergraph; Step S5: Adaptively sparsely encode the plot type on the surveying and mapping enhanced semantic hypergraph to obtain real-time surveying and mapping stream data blocks, and upload them to the land information management platform for information storage tasks.

2. The method for managing construction land surveying and mapping information based on the Internet of Things according to claim 1 is characterized in that: Step S1 is specifically as follows: Step S11: Acquire construction land surveying and mapping data, and perform data preprocessing on the construction land surveying and mapping data to obtain construction land surveying and mapping data to be analyzed; Step S12: performing modal decoupling and feature standardization on the construction land surveying and mapping data to be analyzed to obtain standardized surveying and mapping data; Step S13: performing temporal and spatial feature correlation according to the standardized surveying and mapping data to obtain a surveying and mapping temporal and spatial feature map; Step S14: performing multimodal feature fusion on the surveying and mapping spatiotemporal feature map and constraining the noise surveying and mapping bottleneck to obtain a preliminary perception fusion matrix; Step S15: Calculate the information entropy distribution of the preliminary perception fusion matrix, and perform autoregressive completion and regional boundary optimization on the low-confidence data areas in the information entropy distribution to obtain a multimodal perception matrix for construction land.

3. The method for managing construction land surveying and mapping information based on the Internet of Things according to claim 2 is characterized in that: Step S14 is specifically as follows: Step S141: performing principal component dimensionality reduction based on the surveying and mapping spatiotemporal feature graph, retaining feature components with cumulative contribution rates ≥ 95%, and constructing a surveying and mapping feature subspace; Step S142: performing adaptive weighted attention multimodal feature fusion according to the surveying and mapping feature subspace, dynamically adjusting the feature contribution of different surveying and mapping feature modes in the surveying and mapping feature subspace, and obtaining a multimodal weighted fusion feature matrix; Step S143: Detect low information gain regions based on the multimodal weighted fusion feature matrix, and set the Softmax maximum probability <0.6 to identify low confidence information regions in the multimodal weighted fusion feature matrix; Step S144: combining the low confidence information area and the low information gain area to perform high entropy anomaly detection, identifying the surveying and mapping data noise bottleneck area, and performing adaptive multi-scale fusion to obtain a surveying and mapping bottleneck mask; Step S145: Perform multimodal feature bottleneck constraints on the multimodal weighted fusion feature matrix according to the surveying and mapping bottleneck mask to obtain a preliminary perception fusion matrix.

4. The method for managing construction land surveying and mapping information based on the Internet of Things according to claim 2 is characterized in that: Step S15 is specifically as follows: Step S151: Calculate the information entropy distribution of the preliminary perception fusion matrix, set the entropy segmentation threshold [0.1, 0.5] to divide the high entropy area and the low entropy area, and obtain the mapping entropy partition map; Step S152: Perform low-confidence time series interpolation on the low-entropy area in the mapping entropy partition map, set the confidence range to [0.5, 0.8], and set the spatial consistency coefficient to [0.1, 0.3] to optimize spatial consistency, and obtain a confidence-enhanced mapping matrix; Step S153: Correct the local abnormal surveying and mapping features of the high entropy area in the surveying and mapping entropy partition map to obtain a noise adaptive correction matrix; Step S154: regularizing the region boundary by combining the confidence enhancement mapping matrix and the noise adaptive correction matrix to obtain a boundary optimization mapping matrix; Step S155: Set the consistency evaluation coefficient [0.8, 1.0] to perform probabilistic consistency evaluation on the boundary optimization mapping matrix, and perform adaptive confidence redistribution based on the probabilistic consistency evaluation result to obtain the multimodal perception matrix of the construction land.

5. The method for managing construction land surveying and mapping information based on the Internet of Things according to claim 1, characterized in that: The multi-supervision debiasing of the surveying and mapping data described in step S2 is specifically as follows: Perform latent variable feature decomposition on the multimodal perception matrix of construction land, perform feature orthogonalization, and generate a surveying and mapping latent variable feature set; Unsupervised adversarial correction is performed based on the mapping latent variable feature set to build a cross-device distributed adversarial network, and the cross-device distributed adversarial network is used to perform maximum mean difference alignment on the multimodal perception matrix of the construction land to obtain aligned mapping data; Perform self-supervised abnormal area detection on the aligned surveying and mapping data, and set the abnormal area boundary in combination with the preset confidence threshold to obtain the surveying and mapping abnormal area mask; The aligned surveying and mapping data are combined with the surveying and mapping anomaly area mask to compensate for the low-confidence data area errors and obtain the construction land perception tensor.

6. The method for managing construction land surveying and mapping information based on the Internet of Things according to claim 1, characterized in that: The causal reasoning of organic matter content in step S4 is specifically as follows: Decompose the time series characteristics of the soil evolution trend of the construction land, extract the evolution characteristics of the soil organic matter content, and perform time window normalization to generate the soil evolution characteristic matrix; The Granger causality test of the evolution characteristics is carried out according to the soil evolution characteristic matrix, and the preliminary causal relationship and preliminary causal weight of the soil organic matter content are set based on the causal test results, so as to generate a preliminary soil causal diagram; The preliminary soil causal diagram was analyzed by disturbance variables, redundant factors were eliminated, and the causal effect matrix of soil organic matter content was calculated to construct a causal inference model; The causal inference model is robustly corrected, and the weight distribution in the corrected causal effect matrix is ​​adjusted through the preset causal entropy weight strategy to obtain the optimized soil organic matter causal inference model. Based on the optimized soil organic matter causal inference model, the standardized mapping tensor was reversed to quantify the influencing factors of mapping characteristics.

7. The method for managing construction land surveying and mapping information based on the Internet of Things according to claim 1, characterized in that: The semantic weight of adjusting the standardized mapping tensor described in step S4 is specifically: Construct a preliminary semantic weight matrix based on surveying and mapping feature influencing factors and standardized surveying and mapping tensors; Perform soil characteristic correlation weighted analysis on the preliminary semantic weight matrix, identify the contribution of each surveying and mapping feature in the preliminary semantic weight matrix in reverse deduction, and obtain the deduction contribution of surveying and mapping features; According to the deduced contribution of surveying and mapping features, the feature weights in the preliminary semantic weight matrix are adjusted to obtain an optimized semantic weight matrix; Perform gradient descent processing on the optimized semantic weight matrix, iteratively optimize the weight distribution, and obtain the semantic weight distribution optimization matrix; Performing a stability test on the semantic weight distribution optimization matrix, and adaptively adjusting the semantic weight distribution optimization matrix according to the stability test result to obtain a semantic weight matrix; The semantic weight matrix is ​​mapped to the standardized mapping tensor, the semantic weights of the mapping features are adjusted, and the mapping enhanced semantic hypergraph is obtained.

8. The method for managing construction land surveying and mapping information based on the Internet of Things according to claim 1, characterized in that: Step S5 is specifically as follows: Step S51: clustering surveying and mapping features based on the surveying and mapping enhanced semantic hypergraph to obtain surveying and mapping feature clustering data; Step S52: Obtain land parcel type expert data, perform feature matching on the land parcel type expert data and surveying and mapping feature clustering data, assign land parcel type labels based on feature matching, and obtain land parcel type label data; Step S53: sparsely encode the land parcel type label data to obtain sparsely encoded data; Step S54: acquiring real-time construction land surveying and mapping data through the Internet of Things, and performing real-time land plot type classification and feature coding on the real-time construction land surveying and mapping data based on sparse coding data to obtain real-time surveying and mapping stream data blocks; Step S55: Upload the real-time mapping stream data block to the land information management platform to perform the information storage task.

9. A construction land surveying and mapping information management system based on the Internet of Things, characterized in that: Used to execute the construction land surveying and mapping information management method based on the Internet of Things as claimed in claim 1, the construction land surveying and mapping information management system based on the Internet of Things includes: The data acquisition module is used to obtain construction land surveying and mapping data through the Internet of Things, and to perform collaborative perception fusion on the construction land surveying and mapping data to obtain a multi-modal perception matrix of the construction land; The debiasing module is used to perform multi-supervision debiasing on the multimodal perception matrix of construction land to obtain the perception tensor of construction land, and to perform multi-source consistency correction on the perception tensor of construction land to obtain the standardized mapping tensor; The soil quality evolution trend analysis module is used to obtain the soil quality data of the construction land and analyze the soil quality evolution trend of the construction land based on the soil quality data of the construction land; The causal reasoning module is used to perform causal reasoning on the organic matter content based on the soil evolution trend of the construction land and the standardized surveying and mapping tensor, and obtain the surveying and mapping feature influencing factors; the semantic weight of the standardized surveying and mapping tensor is adjusted based on the surveying and mapping feature influencing factors to obtain the surveying and mapping enhanced semantic hypergraph; The information storage module is used to perform adaptive sparse coding of plot types on the surveying and mapping enhanced semantic hypergraph, obtain real-time surveying and mapping stream data blocks, and upload them to the land information management platform for information storage tasks.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed, the method for managing construction land surveying and mapping information based on the Internet of Things as described in any one of claims 1 to 8 is implemented.

Images