Land utilization dynamic monitoring and analysis method and system based on big data

Abstract

The invention belongs to the technical field of dynamic monitoring, and discloses a land utilization dynamic monitoring and analysis method and system based on big data. Comprising the following steps: collecting optical image data, radar image data and building vector data; registering the optical image data and the radar image data through a sift algorithm; performing qualification analysis on the multi-modal image data and the building vector data after successful registration to obtain a qualified building, obtaining optical image features and radar image features based on the registered multi-modal image data, and generating a correction instruction when a non-qualified building is obtained; calculating multi-modal features based on the optical image features and the radar image features, and training a machine learning model for identifying vegetation and building boundaries in real time based on the multi-modal features; dynamic building changes are monitored based on a machine learning model, building contraction and building expansion are judged, a land utilization change instruction is generated, and the land utilization decision-making efficiency and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the field of dynamic monitoring technology, and more specifically, to a method and system for dynamic monitoring and analysis of land use based on big data. Background Art

[0002] Dynamic monitoring of land use refers to the activity of monitoring and analyzing changes in land use over a certain period of time through remote sensing technology, ground surveys and other means; big data refers to a collection of data that is large in scale, diverse in type, and growing rapidly, and has the characteristics that are difficult to process efficiently by traditional methods. Big data not only involves huge amounts of data, but also involves the entire process of data collection, storage, management and analysis.

[0003] The Chinese patent with the authorization announcement number CN117892149B discloses a landscape change analysis method, device and equipment based on land use data, which uses landscape analysis tools to extract landscape indexes from land use data and classify these indexes through clustering algorithms. The land use data is processed according to the dimension and type of the index to obtain the corresponding landscape index data to identify the change characteristics of the landscape in the target area.

[0004] The Chinese patent with authorization announcement number CN114863289B discloses a dynamic remote sensing monitoring method and system based on land use. By analyzing multi-cycle farming historical data, a crop model associated with the plant growth index is established, the land planting characteristics are theoretically analyzed and the crop growth index is predicted; remote sensing images are used to monitor the real-time growth index of crops; and the planting benefits and economic benefits of cash crops are comprehensively evaluated to determine the overall utilization rate of land in the target area.

[0005] However, the above patents fail to effectively solve the problem of dynamic monitoring in complex surface scenes, such as the junction of buildings and vegetation, and fail to achieve high-precision and dynamic monitoring of land use changes in complex scenes.

[0006] In view of this, the present invention proposes a land use dynamic monitoring and analysis method and system based on big data to solve the above problems. Summary of the invention

[0007] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a method and system for dynamic monitoring and analysis of land use based on big data.

[0008] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a method for dynamic monitoring and analysis of land use based on big data, the method comprising:

[0009] Collect optical image data, radar image data and building vector data;

[0010] Use SIFT algorithm to register optical image data and radar image data;

[0011] The qualified degree of the successfully registered multimodal image data and the building vector data is analyzed. After obtaining the qualified building, the optical image features and radar image features are obtained based on the registered multimodal image data. When an unqualified building is obtained, a correction instruction is generated;

[0012] Calculate multimodal features based on optical image features and radar image features, and train a machine learning model based on multimodal features to identify vegetation and building boundaries in real time;

[0013] Monitor dynamic building changes based on machine learning models, determine building shrinkage and expansion, and generate land use change instructions.

[0014] Preferably, the optical image data captures the boundary between a building and other features; the building vector data refers to the geographic information data of the building represented in a vector format;

[0015] The process of the SIFT algorithm includes extracting SIFT feature key points from optical image data and radar image data; matching feature key points; matching the feature key points extracted from the optical image data and radar image data using nearest neighbor search to obtain the best matching pair; calculating the affine matrix; after matching the best matching pair, calculating the affine matrix by least squares method; applying affine transformation; using affine matrix transformation to align the radar image data to the coordinate system of the optical image data, and marking it as successfully registered multimodal image data.

[0016] Preferably, the conformity analysis process includes setting a building compliance index, the building compliance index is a percentage, which represents the difference between the multimodal image data and the building vector data; comparing the length, width and height of the building vector data with the length, width and height displayed by the multimodal image data in sequence, and taking the sum of the cubes of the difference percentages as the building compliance index;

[0017] Compare and analyze the building compliance index with the preset first-level building compliance threshold. When the building compliance index is less than or equal to the first-level building compliance threshold but greater than the preset second-level building compliance threshold, mark the building as a qualified building; when the building compliance index is greater than the first-level building compliance threshold or less than or equal to the second-level building compliance threshold, mark the building as an unqualified building and generate a correction instruction;

[0018] Preferably, the correction instruction includes sending relevant personnel to conduct an on-site inspection to verify the actual condition of the building. Using CNN to extract the registered radar image features and optical image features respectively;

[0019] The calculation formula of multimodal features is:

[0020] Q=W Q F opt ;

[0021] K=W K F rad ;

[0022] V=W V F rad ;

[0023] In the formula, F opt Indicates optical image characteristics; F rad represents the radar image feature; the query Q is the optical image feature projection, the key K and value V are the feature projection of the radar image;

[0024] W Q , W K , W V is the weight matrix used for linear transformation;

[0025] Score is the weight matrix; Softmax is the normalization function;

[0026] F fused =Score·V;F fused It is a multimodal feature;

[0027] Preferably, the training process of the machine learning model includes:

[0028] The corresponding real semantics in the multimodal features are used as the labels corresponding to the multimodal features, each group of multimodal features and the corresponding labels are constructed as a sample, and multiple samples are collected to construct a data set; the data set is divided into a training set, a validation set and a test set, wherein the training set accounts for 60% of the data set, and the validation set and the test set each account for 20% of the data set; the meaning of the semantics is the real object corresponding to the multimodal feature map,

[0029] The training set is used as the input of the machine learning model, and the machine learning model uses the semantics of the real-time multimodal features as the output; the semantics corresponding to a set of real-time multimodal features is used as the prediction target, and minimizing the loss function value of the machine learning model is used as the training target; when the loss function value of the machine learning model is less than or equal to the preset target loss value, the training is stopped;

[0030] The machine learning model loss function is the mean square error; the mean square error is calculated by Minimize to train the model; in the loss function, MSE is the loss function value, i is the multimodal feature group number; u is the number of multimodal feature groups; y i is the label corresponding to the i-th group of multimodal features, The semantics predicted for the i-th group of multimodal features;

[0031] The machine learning model is a deep neural network model;

[0032] The calculation method of the hidden layer H of the deep neural network model includes:

[0033] H=G[b i +w i s(b i-1 +w i-1 x)]; where H is the output of the hidden layer; G is the activation function; b i is the bias term of the hidden layer; w i is the weight of the hidden layer; s is the activation function; b i-1 is the bias term of the previous layer; w i-1 is the weight of the previous layer; x is the input training set;

[0034] The deep neural network model consists of 4 fully connected layers. The first 3 hidden layers contain 32, 64 and 32 neurons respectively. There is 1 SoftMax classification layer with 4 neurons. The neurons are connected to each other through weights and activated by ReLU activation function.

[0035] Preferably, when the machine learning model identifies the building boundary, the number of building boundary pixel points is obtained; the multimodal features with the same historical geographical location are obtained to obtain the number of historical building boundary pixel points; the difference between the number of building boundary pixel points and the number of historical building boundary pixel points is compared with a preset first-level boundary change threshold, and when the difference is greater than or equal to the first-level boundary change threshold, the building expansion is marked; when the difference is less than or equal to the preset second-level boundary change threshold, it is marked as building contraction; the second-level boundary change threshold is less than the first-level boundary change threshold; when building expansion and building contraction are marked, a land use change instruction is generated;

[0036] The land use change instructions include arranging staff to check whether the building land has expanded or shrunk illegally, and whether the urban greening meets the standards after the building changes.

[0037] A land use dynamic monitoring and analysis system based on big data, the system comprising:

[0038] Data collection module, which collects optical image data, radar image data and building vector data;

[0039] The preprocessing module registers the optical image data and radar image data through the sift algorithm;

[0040] The qualification analysis module performs qualification analysis on the successfully registered multimodal image data and building vector data. After obtaining qualified buildings, the optical image features and radar image features are obtained based on the registered multimodal image data. When unqualified buildings are obtained, correction instructions are generated;

[0041] Model training module, which calculates multimodal features based on optical image features and radar image features, and trains a machine learning model for real-time recognition of vegetation and building boundaries based on multimodal features;

[0042] The decision-making module monitors dynamic building changes based on machine learning models, determines building contraction and expansion, and generates land use change instructions.

[0043] An electronic device includes: a processor and a memory, wherein the memory stores a computer program that can be called by the processor; the processor executes the above-mentioned big data-based dynamic land use monitoring and analysis method by calling the computer program stored in the memory.

[0044] A computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned method for dynamic monitoring and analysis of land use based on big data.

[0045] The technical effects and advantages of the method and system for dynamic monitoring and analysis of land use based on big data of the present invention: This application realizes accurate identification and dynamic monitoring of the boundaries between buildings and vegetation through the mutual coordination of multimodal image data (optical images, radar images) and building vector data. Collect optical image and radar image data and perform geometric correction and SIFT registration to ensure the spatial consistency of multimodal data; use CNN to extract optical and radar image features respectively, and fuse the two modal data through the attention mechanism to generate a more comprehensive target representation; perform qualification analysis on the registered multimodal images and building vector data, screen compliant buildings for subsequent feature extraction and model training, train deep neural network models based on multimodal features, realize real-time semantic recognition of the boundaries between buildings and vegetation, and dynamically monitor building changes through historical comparisons to generate land use change instructions.

[0046] This application combines the advantages of optical images and radar images to achieve high-precision identification and dynamic change monitoring of building and vegetation boundaries in complex scenes. The rich spectral information of optical images and the geometric features of radar images complement each other, improving the accuracy of boundary detection. The fusion of multi-source data improves the semantic recognition accuracy and robustness of the model in complex remote sensing scenes, and realizes real-time monitoring of building boundary changes. The conformity analysis ensures data quality and reduces error accumulation, so that the solution has a wide range of application value, such as urban planning, ecological monitoring and land use supervision. This application is helpful in urban planning, ecological monitoring and other fields, ensures the compliance of urban greening and land use, improves resource management efficiency, and provides decision makers with a scientific basis for land use changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic diagram of the land use dynamic monitoring and analysis method based on big data of the present invention;

[0048] Figure 2 It is a schematic diagram of the land utilization dynamic monitoring and analysis system based on big data of the present invention;

[0049] Figure 3 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application;

[0050] Figure 4 It is a schematic diagram of the structure of a computer-readable storage medium provided by an embodiment of the present application. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] Example 1

[0053] See also Figure 1 As shown, the land use dynamic monitoring and analysis method based on big data described in this embodiment includes:

[0054] Collect optical image data; in addition to the visible light band, optical image data also includes near-infrared and short-wave infrared bands based on traditional photos. These bands are beyond the visual range of the human eye, but can provide more spectral information about surface objects.

[0055] Optical image data captures the boundary between buildings and other objects; the red edge band and near-infrared band of optical images can highlight the difference in reflection between buildings and other objects, especially in urban areas. In addition, the near-infrared band of optical image data is extremely sensitive to vegetation and can accurately distinguish healthy vegetation from non-vegetated areas. Optical image data is acquired through the Gaofen series of satellites or in real time through the Sentinel-2 satellite.

[0056] The Gaofen series of satellites are a series of earth observation satellites that cover multiple spectral bands from visible light to infrared, and can meet the needs of different application fields. The data from these satellites play an important role in agriculture, forestry, environmental protection, urban planning, disaster monitoring, etc. For example, Gaofen-5 is a hyperspectral comprehensive observation satellite launched in 2018, mainly used for environmental monitoring and resource survey.

[0057] Sentinel-2 is an Earth observation satellite launched by the European Space Agency as part of the Copernicus program. Sentinel-2 provides multispectral, high-resolution image data for monitoring changes on the Earth's surface.

[0058] Collect radar image data; radar images capture the geometric and roughness characteristics of the surface by actively transmitting microwave signals and receiving echo signals reflected from the surface. Compared with optical images, radar images are not affected by weather and lighting conditions.

[0059] Radar image data can be obtained from the Tianhui-1 (TH-1) satellite series, which is a remote sensing satellite with both optical and radar imaging capabilities. Their radar payload can provide high-resolution radar images for terrain mapping, resource surveys, and environmental monitoring.

[0060] Collect building vector data; building vector data refers to the geographic information data of buildings represented in vector format. In the geographic information system, vector data is a data type used to represent geographic features. It accurately describes the location, shape and size of objects through geometric shapes such as points, lines and polygons. Building vector data is obtained by relevant urban planning and design departments.

[0061] Collecting and expanding a variety of high-quality auxiliary data can ensure comprehensive coverage of areas where buildings and vegetation intersect, while generating richer training data to improve the change detection capabilities of subsequent models.

[0062] Geometrically calibrate optical image data, radar image data and building vector data;

[0063] Different data sources may have inconsistent spatial positions during the acquisition process due to imaging angle, earth curvature, satellite orbit parameters, etc. The goal of geometric correction is to align these images to the same coordinate system to ensure spatial consistency in subsequent analysis.

[0064] Use SIFT algorithm to register optical image data and radar image data;

[0065] SIFT is an algorithm for extracting local feature points of images. It can identify key corner points, edges and other features in images and is invariant to scaling and rotation.

[0066] Extract SIFT feature key points from optical image data and radar image data, and set up a descriptor for each key point;

[0067] Matching feature key points; using nearest neighbor search to match the feature key points extracted from the optical image data and radar image data to obtain the best matching pair.

[0068] Calculate the geometric transformation parameters; after matching the best matching pair, calculate the affine matrix, which describes how to map points in the radar image data to the coordinate system of the optical image data. Specifically, the affine matrix can be calculated by the least squares method to reduce the influence of external points and improve the accuracy of the transformation matrix.

[0069] Apply affine transformation; use affine transformation to align radar image data to the coordinate system of optical image data, and mark it as successfully registered multimodal image data. Affine transformation can correct geometric distortions such as translation, rotation, scaling and shearing.

[0070] The use of SIFT matching algorithm ensures the spatial consistency of data, lays the foundation for multimodal data fusion, improves the reliability of the model in the feature extraction stage, and avoids errors caused by coordinate differences.

[0071] After successful registration, the multimodal image data and building vector data are analyzed for conformity; the conformity analysis process includes:

[0072] Set the building compliance index, which is a percentage that indicates the difference between the multimodal image data and the building vector data. Specifically, the length, width and height of the building vector data are compared with the length, width and height shown by the multimodal image data in turn, and the sum of the cubes of the difference percentages is used as the building compliance index.

[0073] For example, the boundary of a building in the building vector is 200m long, 150m wide, and 10m high, but the multimodal image data shows that the length is 210m, the width is 160m, and the height is 11m. The calculation formula for the building compliance index is ((210-200) / 200)3 +((160-150) / 150) 3 +((11-10) / 10) 3 ;

[0074] Compare and analyze the building compliance index with the preset first-level building compliance threshold. When the building compliance index is less than or equal to the first-level building compliance threshold but greater than the preset second-level building compliance threshold, the building is marked as a qualified building. At this time, the boundary line of the building is also relatively accurate, which is suitable for subsequent processing of building boundaries and vegetation boundaries; when the building compliance index is greater than the first-level building compliance threshold or less than or equal to the second-level building compliance threshold, the building is marked as an unqualified building and a correction instruction is generated. An unqualified building means that the construction of this building exceeds the original plan or is far below the original plan; at this time, the ground environment of the building boundary changes a lot, and the multimodal image data is quite different from the originally planned building vector data.

[0075] By setting compliance indicators, it is possible to effectively monitor whether the construction project is implemented according to the plan, and timely discover situations that exceed the plan or fail to meet the planning requirements, thereby promoting compliance inspections;

[0076] Corrective instructions include sending relevant personnel to conduct on-site inspections to verify the actual conditions of the building.

[0077] The main goal of registration is to ensure that optical images and radar images are aligned in the same geographic space, but this does not mean that the two have the same feature expression. Optical images mainly provide spectral features such as reflectivity and texture. Radar images emphasize the backscattering characteristics such as the structural characteristics and surface roughness of the target object. Even if the two are aligned in space, they are still data from different sources and with different feature expressions. Therefore, the registered multimodal image data needs to be further fused to fully explore the complementary information of the two modalities.

[0078] It should be noted that the multimodal features are obtained only when the building is marked as a qualified building. Otherwise, the building itself has errors, and the accuracy of the surrounding vegetation environment changes is greatly reduced, which is different from the planned building.

[0079] Based on the registered multimodal image data, optical image features and radar image features are obtained; CNN is used to extract the registered radar image features and optical image features respectively; Convolutional neural network (CNN) can extract features because it simulates the local perception and hierarchical processing mechanism of the human visual system, uses learnable convolution kernels to slide on the image to capture local features, gradually learns the image through weight sharing and multi-layer stacking, and the introduction of nonlinear activation functions and pooling layers enhances the model's ability to extract features. Multimodal features are calculated based on optical image features and radar image features.

[0080] The calculation formula of multimodal features is:

[0081] Q=W Q F opt ;

[0082] K=W K F rad ;

[0083] V=W V F rad ;

[0084] Where F opt Indicates optical image characteristics; F rad Represents the radar image features; query Q is the optical image feature projection, which is used to indicate what information it needs to learn from the radar image. Key K and value V are the feature projections of the radar image; key K is used to represent the feature content of the radar image. Value V is used to represent the specific feature value of the radar image.

[0085] W Q , W K , W V is the weight matrix used for linear transformation;

[0086] Score is the weight matrix; Softmax is the normalization function

[0087] F fused =Score·V;F fused It is a multimodal feature;

[0088] The two modal data represent the same target differently, but they complement each other. The boundary of a building may depend on texture and color changes in optical images, but on high reflection intensity in radar images. This attention mechanism can select the relevant features required for optical images from radar images, learn the association between the two modalities, and effectively combine the two to obtain a more comprehensive target representation. This is conducive to the effective recognition of boundaries.

[0089] A machine learning model for real-time recognition of vegetation and building boundaries based on multimodal feature training;

[0090] The training process of the machine learning model includes:

[0091] The corresponding real semantics in the multimodal features are used as the labels corresponding to the multimodal features, each group of multimodal features and the corresponding labels are constructed into a sample, and multiple samples are collected to construct a data set; the data set is divided into a training set, a validation set and a test set, wherein the training set accounts for 60% of the data set, and the validation set and the test set each account for 20% of the data set; the meaning of the semantics is the real object corresponding to the multimodal feature map, such as vegetation or building boundaries, and the multimodal features are annotated by relevant engineers according to actual conditions.

[0092] The training set is used as the input of the machine learning model, and the machine learning model uses the semantics of the real-time multimodal features as the output; the semantics corresponding to a set of real-time multimodal features is used as the prediction target, and minimizing the loss function value of the machine learning model is used as the training target; when the loss function value of the machine learning model is less than or equal to the preset target loss value, the training is stopped;

[0093] The machine learning model loss function is the mean square error; the mean square error is calculated by Minimize to train the model; in the loss function, MSE is the loss function value, i is the multimodal feature group number; u is the number of multimodal feature groups; y i is the label corresponding to the i-th group of multimodal features, The semantics predicted for the i-th group of multimodal features;

[0094] The machine learning model is a deep neural network model;

[0095] The calculation method of the hidden layer H of the deep neural network model includes:

[0096] H=G[b i +w i s(b i-1 +w i-1 x)]; where H is the output of the hidden layer; G is the activation function; b i is the bias term of the hidden layer; w i is the weight of the hidden layer; s is the activation function; b i-1 is the bias term of the previous layer; w i-1 is the weight of the previous layer; x is the input training set.

[0097] The deep neural network model consists of 4 fully connected layers. The first 3 hidden layers contain 32, 64 and 32 neurons respectively. There is 1 SoftMax classification layer with 4 neurons. The neurons are connected to each other through weights and activated by ReLU activation function.

[0098] In this way, the expressive power of the model is enhanced, while ensuring real-time performance and application scalability. This design is particularly suitable for complex remote sensing scenarios, and can achieve high-precision and high-robust semantic recognition in the boundary areas of buildings, vegetation and other categories, which has important application value in urban planning, ecological monitoring and other fields.

[0099] Monitor dynamic building changes based on machine learning models; when the machine learning model identifies the building boundary, obtain the number of building boundary pixels; specifically implemented by PYTHON code; obtain multimodal features with the same historical geographical location, and obtain the number of historical building boundary pixels; compare the difference between the number of building boundary pixels and the number of historical building boundary pixels with the preset first-level boundary change threshold, when the difference is greater than or equal to the first-level boundary change threshold, mark the building expansion; when the difference is less than or equal to the preset second-level boundary change threshold, mark it as building contraction; the second-level boundary change threshold is less than the first-level boundary change threshold; when building expansion and building contraction are marked, generate land use change instructions.

[0100] The land use change instructions include arranging staff to check whether the building land has expanded or shrunk illegally, and whether the urban greening meets the standards after the building changes.

[0101] This application realizes accurate identification and dynamic monitoring of the boundaries between buildings and vegetation through the mutual coordination of multimodal image data (optical images, radar images) and building vector data. Collect optical image and radar image data and perform geometric correction and SIFT registration to ensure the spatial consistency of multimodal data; use CNN to extract optical and radar image features respectively, and fuse the two modal data through the attention mechanism to generate a more comprehensive target representation; perform qualification analysis on the registered multimodal images and building vector data, select compliant buildings for subsequent feature extraction and model training, train deep neural network models based on multimodal features, realize real-time semantic recognition of the boundaries between buildings and vegetation, and dynamically monitor building changes through historical comparisons to generate land use change instructions.

[0102] This application combines the advantages of optical images and radar images to achieve high-precision identification and dynamic change monitoring of building and vegetation boundaries in complex scenes. The rich spectral information of optical images and the geometric features of radar images complement each other, improving the accuracy of boundary detection. The fusion of multi-source data improves the semantic recognition accuracy and robustness of the model in complex remote sensing scenes, and realizes real-time monitoring of building boundary changes. The conformity analysis ensures data quality and reduces error accumulation, so that the solution has a wide range of application value, such as urban planning, ecological monitoring and land use supervision. This application is helpful in urban planning, ecological monitoring and other fields, ensuring the compliance of urban greening and land use, improving resource management efficiency, and providing decision makers with a scientific basis for land use changes.

[0103] Example 2

[0104] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, and a method and system for dynamic monitoring and analysis of land use based on big data are provided, wherein the system comprises: a data collection module for collecting optical image data, radar image data and building vector data;

[0105] The preprocessing module registers the optical image data and radar image data through the SIFT algorithm;

[0106] The qualification analysis module performs qualification analysis on the successfully registered multimodal image data and building vector data. After obtaining qualified buildings, the optical image features and radar image features are obtained based on the registered multimodal image data. When unqualified buildings are obtained, correction instructions are generated;

[0107] Model training module, which calculates multimodal features based on optical image features and radar image features, and trains a machine learning model for real-time recognition of vegetation and building boundaries based on multimodal features;

[0108] The decision-making module monitors dynamic building changes based on machine learning models, determines building contraction and expansion, and generates land use change instructions.

[0109] The modules are connected via wired and / or wireless networks.

[0110] Example 3

[0111] See also Figure 3 As shown, according to another aspect of the present application, an electronic device 500 is also provided. The electronic device 500 may include one or more processors and one or more memories. The memories store computer readable codes, and when the computer readable codes are executed by one or more processors, the land use dynamic monitoring and analysis method and system based on big data as described above can be executed.

[0112] The method or system according to the embodiment of the present application can also be Figure 3 The electronic device shown in the figure is implemented by the electronic device architecture. Figure 3As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a read-only memory (ROM) 503, a random access memory (RAM) 504, a communication port 505 connected to a network, an input / output component 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, may store the land use dynamic monitoring and analysis method and system based on big data provided by the present application. Furthermore, the electronic device 500 may also include a user interface 508. Of course, Figure 3 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 3 One or more components of an electronic device are shown.

[0113] Example 4

[0114] See also Figure 4 As shown, a computer-readable storage medium 600 according to an embodiment of the present application is shown. Computer-readable instructions are stored on the computer-readable storage medium 600. When the computer-readable instructions are executed by the processor, the land utilization dynamic monitoring and analysis method and system based on big data according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium 600 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache). Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0115] In addition, according to the embodiments of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided by the present application. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0116] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center through a wired network or a wireless network. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0117] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0118] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0119] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a method and system for dynamic monitoring and analysis of land use based on big data. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0120] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0121] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0122] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0123] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for dynamic monitoring and analysis of land use based on big data, characterized in that: include: Collect optical image data, radar image data and building vector data; Use SIFT algorithm to register optical image data and radar image data; The qualified degree of the successfully registered multimodal image data and the building vector data is analyzed. After obtaining the qualified building, the optical image features and radar image features are obtained based on the registered multimodal image data. When an unqualified building is obtained, a correction instruction is generated; Calculate multimodal features based on optical image features and radar image features, and train a machine learning model based on multimodal features to identify vegetation and building boundaries in real time; Monitor dynamic building changes based on machine learning models, determine building shrinkage and expansion, and generate land use change instructions.

2. The method for dynamic monitoring and analysis of land use based on big data according to claim 1 is characterized in that: The optical image data captures the boundary between buildings and other landforms; building vector data refers to the geographic information data of buildings expressed in vector format; The process of the sift algorithm includes extracting SIFT feature key points from optical image data and radar image data; Matching feature key points; The characteristic key points extracted from the optical image data and the radar image data are matched using the nearest neighbor search to obtain the best matching pair; the affine matrix is ​​calculated; after the best matching pair is obtained, the affine matrix is ​​calculated by the least squares method; the affine transformation is applied; the radar image data is aligned to the coordinate system of the optical image data using the affine matrix transformation, and marked as the multimodal image data with successful registration.

3. The method for dynamic monitoring and analysis of land use based on big data according to claim 2 is characterized in that: The conformity analysis process includes setting a building compliance index, which is a percentage that represents the difference between the multimodal image data and the building vector data; The length, width and height of the building vector data are compared with the length, width and height shown by the multimodal image data, and the sum of the cubes of the difference percentages is used as an indicator of the building's compliance level; Compare and analyze the building compliance index with the preset first-level building compliance threshold, and when the building compliance index is less than or equal to the first-level building compliance threshold but greater than the preset second-level building compliance threshold, mark the building as a qualified building; When the building compliance index is greater than the first-level building compliance threshold or less than or equal to the second-level building compliance threshold, the building is marked as a non-qualified building and a correction instruction is generated.

4. The method for dynamic monitoring and analysis of land use based on big data according to claim 3 is characterized in that: Correction instructions include sending relevant personnel to conduct field inspections to verify the actual conditions of the building. Use CNN to extract the registered radar image features and optical image features respectively; The calculation formula of multimodal features is: Q=W Q F opt ; K=W K F rad ; V=W V F rad ; In the formula, F opt Indicates optical image characteristics; F rad Indicates radar image features; The query Q is the feature projection of the optical image, and the key K and value V are the feature projections of the radar image; W Q , W K , W V is the weight matrix used for linear transformation; Score is the weight matrix; Softmax is the normalization function; F fused =Score·V;F fused It is a multimodal feature.

5. The method for dynamic monitoring and analysis of land use based on big data according to claim 4 is characterized in that: The training process of the machine learning model includes: The corresponding real semantics in the multimodal features are used as the labels corresponding to the multimodal features, each group of multimodal features and the corresponding labels are constructed as a sample, and multiple samples are collected to construct a data set; the data set is divided into a training set, a validation set and a test set, wherein the training set accounts for 60% of the data set, and the validation set and the test set each account for 20% of the data set; the meaning of the semantics is the real object corresponding to the multimodal feature map, The training set is used as the input of the machine learning model, and the machine learning model uses the semantics of the real-time multimodal features as the output; the semantics corresponding to a set of real-time multimodal features is used as the prediction target, and minimizing the loss function value of the machine learning model is used as the training target; when the loss function value of the machine learning model is less than or equal to the preset target loss value, the training is stopped; The machine learning model loss function is the mean square error; the mean square error is calculated by Minimize to train the model; in the loss function, MSE is the loss function value, i is the multimodal feature group number; u is the number of multimodal feature groups; y i is the label corresponding to the i-th group of multimodal features, The semantics predicted for the i-th group of multimodal features; The machine learning model is a deep neural network model; The calculation method of the hidden layer H of the deep neural network model includes: H=G[b i +w i s(b i-1 +w i-1 x)]; where H is the output of the hidden layer; G is the activation function; b i is the bias term of the hidden layer; w i is the weight of the hidden layer; s is the activation function; b i-1 is the bias term of the previous layer; w i-1 is the weight of the previous layer; x is the input training set; The deep neural network model consists of 4 fully connected layers. The first 3 hidden layers contain 32, 64 and 32 neurons respectively. There is 1 SoftMax classification layer with 4 neurons. The neurons are connected to each other through weights and activated by ReLU activation function.

6. The method for dynamic monitoring and analysis of land use based on big data according to claim 5 is characterized in that: When the machine learning model identifies the building boundary, the number of building boundary pixel points is obtained; the multimodal features with the same historical geographical location are obtained to obtain the number of historical building boundary pixel points; the difference between the number of building boundary pixel points and the number of historical building boundary pixel points is compared with the preset first-level boundary change threshold, and when the difference is greater than or equal to the first-level boundary change threshold, the building expansion is marked; When the difference is less than or equal to the preset secondary boundary change threshold, it is marked as building contraction; when the secondary boundary change threshold is less than the primary boundary change threshold; when building expansion and building contraction are marked, a land use change instruction is generated; The land use change instructions include arranging staff to check whether the building land has expanded or shrunk illegally, and whether the urban greening meets the standards after the building changes.

7. The land use dynamic monitoring and analysis system based on big data is characterized by: Data collection module, which collects optical image data, radar image data and building vector data; The preprocessing module registers the optical image data and radar image data through the SIFT algorithm; The qualification analysis module performs qualification analysis on the successfully registered multimodal image data and building vector data. After obtaining qualified buildings, the optical image features and radar image features are obtained based on the registered multimodal image data. When unqualified buildings are obtained, correction instructions are generated; Model training module, which calculates multimodal features based on optical image features and radar image features, and trains a machine learning model for real-time recognition of vegetation and building boundaries based on multimodal features; The decision-making module monitors dynamic building changes based on machine learning models, determines building contraction and expansion, and generates land use change instructions.

8. An electronic device, characterized in that include: A processor and a memory, wherein the memory stores a computer program that can be called by the processor; The processor executes the land use dynamic monitoring and analysis method based on big data as described in any one of claims 1 to 6 by calling the computer program stored in the memory.

9. A computer-readable storage medium, characterized in that: Instructions are stored, and when the instructions are executed on a computer, the computer executes the land use dynamic monitoring and analysis method based on big data as described in any one of claims 1 to 6.

Images