River natural shoreline intelligent identification method and system based on multi-source remote sensing data

Abstract

The invention discloses a river natural shoreline intelligent identification method and system based on multi-source remote sensing data, relates to the technical field of geographic information processing, environment remote sensing monitoring and smart city ecological planning, and solves the problems of low efficiency and insufficient precision of traditional identification. Comprising the following steps: acquiring data such as RGB / multispectral images with the resolution of less than or equal to 1m and laser radar point clouds; carrying out radiation correction, geometric registration and NDWI segmentation preprocessing; initially extracting a continuous shoreline by using a Canny operator and topological optimization; multi-dimensional features such as curvature, gradient and NDVI are calculated; a natural shoreline probability graph is output through a random forest classifier with the tree number being 500 and the maximum depth being 10, a distribution graph is generated after morphological filtering optimization, and related data are counted. The system comprises a remote sensing data server and other hardware and corresponding software modules. According to the invention, high-precision identification of the natural shoreline is realized, and support is provided for urban river ecological protection and shoreline planning.

Description

Technical Field

[0001] This invention relates to the fields of geographic information processing, environmental remote sensing monitoring, and smart city ecological planning, specifically to a method and system for intelligent identification of natural river shorelines based on multi-source remote sensing data. Background Technology

[0002] As a key intersection of aquatic and terrestrial ecosystems, the classification of urban river shorelines (natural and artificial shorelines) is crucial for ecological protection, water resource management, and urban planning. With the acceleration of urbanization, the proportion of artificial shorelines continues to rise, making the accurate identification and protection of natural shorelines an urgent need for ecological governance.

[0003] Traditional shoreline identification methods often rely on on-site surveys or single remote sensing data, which suffer from problems such as low efficiency, limited coverage, and significant interference from topography, making it difficult to meet the needs of large-scale, high-precision urban river shoreline monitoring.

[0004] Existing remote sensing identification technologies often suffer from insufficient differentiation between natural and artificial shorelines due to limited data dimensions and incomplete feature extraction, leading to misjudgments and omissions, especially in complex urban river sections where identification accuracy is difficult to guarantee.

[0005] Therefore, a method and system for intelligent identification of natural river shorelines based on multi-source remote sensing data are proposed to solve the above problems. Summary of the Invention

[0006] In view of this, the technical problem to be solved by the present invention is to propose a method and system for intelligent identification of natural river shorelines based on multi-source remote sensing data, so as to solve the problems in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method and system for intelligent identification of natural river shorelines based on multi-source remote sensing data, comprising: S1, Multi-source data acquisition: Acquire high-resolution remote sensing images, lidar point cloud data, and auxiliary data. The high-resolution remote sensing images are RGB and multispectral images with a resolution ≤1m. The lidar point cloud data is used to generate digital elevation models (DEM) and digital surface models (DSM). The auxiliary data includes historical land use maps and hydrogeological data. S2, Data Preprocessing: Radiometric correction, geometric registration, and water body mask extraction are performed sequentially on the acquired multi-source data; the radiometric correction is used to eliminate the influence of atmospheric scattering; the geometric registration aligns the image with the digital elevation model (DEM), and the registration error is <1 pixel; the water body mask extraction uses the normalized difference water index (NDWI) to segment the main body of the river. S3, Initial shoreline extraction: Edge detection is performed using the Canny operator to identify the water-land boundary, and then topology optimization is used to remove broken segments to generate a continuous shoreline; S4, Multi-dimensional Feature Calculation: Geometric morphological features, vegetation cover features, and texture features are calculated separately; the geometric morphological features include curvature and slope, wherein curvature is calculated using the sliding window method, and the calculation formula is:

[0008] In the formula, Represented as curvature value; It is expressed as the radius of curvature of the curve at that point, and the curvature value of a natural shoreline is >0.15; The slope is calculated based on the DEM using the maximum gradient method. The specific formula is as follows: S

[0009] In the formula, S is the slope value. , These are the elevation change rates in the x and y directions, respectively, and the slope range of the natural shoreline is 5°-45°. S5, Classifier Training and Recognition: Construct a training set containing natural and artificial shoreline samples. After fusing the geometric morphological features, vegetation cover features, and texture features calculated in step 4, input them into a random forest classifier and output a natural shoreline probability map with a probability value range of 0-1.0. The random forest classifier has 500 trees and a maximum depth of 10. S6, Post-processing optimization: Morphological filtering is used to remove small noise patches, and adjacent natural shoreline segments are merged through spatial continuity constraints; S7, Results Visualization Output: Generate a shoreline type distribution map, clarify the distribution range of natural and artificial shorelines, and statistically analyze the proportion and segment length of natural shorelines.

[0010] Preferably, the width of the shoreline buffer zone in S1 is set to 20-30m, and the vegetation cover feature is extracted using NDVI enhancement to obtain the vegetation index within the shoreline buffer zone. The calculation formula is as follows:

[0011] In the formula, For near-infrared reflectivity, Reflectivity in the red light band, natural coastline Value > 0.3; The texture features are calculated based on the GLCM gray-level co-occurrence matrix, using entropy and contrast values. The entropy calculation formula is as follows:

[0012] In the formula, Let i be the probability of elements i and j in the gray-level co-occurrence matrix, and let the entropy of the natural shoreline be > 2.0. The formula for calculating contrast is:

[0013] As a preferred embodiment, the training set described in S1 is obtained by manually annotating urban river shoreline samples, with the samples covering shoreline areas of different landform types, vegetation coverage, and artificial modification intensity.

[0014] A smart river shoreline identification system based on multi-source remote sensing data includes hardware components and software modules. The hardware components include a remote sensing data server, a GPU computing node, and a terminal display device. The software modules include a data acquisition module, a data preprocessing module, a shoreline initial extraction module, a multi-dimensional feature calculation module, a classification and identification module, a post-processing optimization module, and a result output module. The data acquisition module is used to collect high-resolution remote sensing images, lidar point cloud data, and auxiliary data. The high-resolution remote sensing images are RGB / multispectral images with a resolution ≤1m. The lidar point cloud data is used to generate DEM and DSM. The auxiliary data includes historical land use maps and hydrogeological data. The data preprocessing module is used to perform radiometric correction, geometric registration, and water body mask extraction on the acquired multi-source data. The radiometric correction eliminates the influence of atmospheric scattering, the geometric registration aligns the image with the DEM with an error of <1 pixel, and the water body mask extraction uses NDWI exponential segmentation of the river body. The initial shoreline extraction module uses the Canny operator to perform edge detection and identify the water-land boundary, and generates a continuous shoreline through topology optimization; The multi-dimensional feature calculation module is used to calculate geometric features, vegetation cover features, and texture features. The geometric features include curvature and slope. Curvature is calculated using the sliding window method, and slope is calculated based on DEM. The vegetation cover feature is the NDVI value within the shoreline buffer. The texture feature is the entropy value of the GLCM gray-level co-occurrence matrix. The classification and recognition module includes a training set construction unit and a random forest classifier. The training set construction unit is used to store manually labeled natural / artificial shoreline samples. The random forest classifier receives multi-dimensional feature fusion data and outputs a natural shoreline probability map. The random forest classifier has 500 trees and a maximum depth of 10. The post-processing optimization module uses morphological filtering to remove noise and spatial continuity constraints to merge adjacent natural shoreline segments. The result output module is used to generate a shoreline type distribution map and statistically analyze the proportion and segment length of natural shorelines, and display it on the terminal display device.

[0015] Preferably, the remote sensing data server is used to store high-resolution remote sensing images and lidar point cloud data acquired by satellites and UAVs, and the GPU computing node is used to accelerate the multi-dimensional feature extraction and random forest classifier operation.

[0016] Preferably, in the multi-dimensional feature calculation module, the curvature threshold of the natural shoreline is >0.15, the slope range is 5°-45°, the NDVI threshold is >0.3, and the entropy threshold is >2.0.

[0017] Preferably, the width of the shoreline buffer zone is set to 30m, and the morphological filtering uses a combination of opening and closing operations to remove small noise patches.

[0018] Compared with existing technologies, the intelligent identification method and system for natural river shorelines based on multi-source remote sensing data provided by this invention has the following advantages: (1) Multi-source data fusion improves the reliability of identification: integrate high-resolution RGB / multispectral images, lidar point cloud data and historical land use maps and other auxiliary data to make up for the information shortcomings of a single data source. At the same time, through a strict data preprocessing process, radiometric correction, geometric registration, etc., it provides high-quality data support for subsequent identification.

[0019] (2) Multi-dimensional feature construction achieves accurate differentiation: shoreline features are extracted from three core dimensions: geometric shape including curvature, slope, vegetation cover based on NDVI value, and texture features including entropy value and contrast. This comprehensively depicts the inherent attributes of natural shorelines, significantly improves the differentiation from artificial shorelines, and reduces misjudgment caused by single features.

[0020] (3) Intelligent classification and optimization process ensures recognition accuracy: A random forest classifier with 500 trees and a maximum depth of 10 is used, combined with training of diverse manually labeled samples to output a high-confidence natural shoreline probability map; subsequently, through morphological filtering and spatial continuity constraint optimization, noise is effectively removed and broken shorelines are merged, and the overall recognition accuracy can reach more than 90%.

[0021] (4) Efficient visualization and quantitative output to support decision-making: It can not only generate clear shoreline type distribution maps, but also accurately count key data such as the proportion of natural shorelines and segment lengths. Furthermore, it accelerates computation through GPU computing nodes, taking into account both recognition efficiency and the practicality of results, and provides direct data support for urban river ecological protection and shoreline planning and management. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the process for intelligent identification of natural river shorelines based on multi-source remote sensing data according to the present invention. Figure 2This is a schematic diagram of the intelligent river shoreline identification system based on multi-source remote sensing data according to the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0024] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0025] For an example, please refer to... Figures 1 to 2 As shown: To address the problems mentioned in the technical solutions, this application provides an intelligent method for identifying natural river shorelines based on multi-source remote sensing data. The specific steps are as follows: Step 1: Acquisition of multi-source data and parameter setting; Data acquisition: Download the aforementioned high-resolution remote sensing images and airborne LiDAR point cloud data through the remote sensing data server, import historical land use maps and hydrogeological data, and complete the multi-source data collection; Buffer zone setting: Based on the shoreline characteristics of the study area, the width of the shoreline buffer zone is set to 30m (which meets the setting requirement of 20-30m).

[0026] Step 2, data preprocessing; Radiometric correction: ENVI software was used to perform radiometric correction on WorldView-3 imagery. The atmospheric correction module was used to eliminate the effects of atmospheric scattering and absorption, and to obtain the true surface reflectance data. Geometric registration: Using a 1m resolution DEM as a reference, geometric registration is performed on the corrected remote sensing image. A quadratic polynomial registration model is adopted, and the registration error is controlled within 0.8 pixels (meeting the requirement of <1 pixel). Water body mask extraction: Calculate the Normalized Differential Water Index (NDWI), set a threshold of 0.1, segment the main river area, generate a water body mask layer, and exclude interference from non-water body areas.

[0027] Step 3, Initial shoreline extraction Edge detection: In ArcGIS software, the Canny operator is called to perform edge detection on the water mask layer, identify the water-land boundary, and obtain the initial shoreline segment; Topology optimization: Using a topology repair tool, broken segments in the initial shoreline are removed (segments with a length of <5m are considered broken segments), and adjacent valid segments are connected to generate a continuous initial shoreline.

[0028] Step 4: Multi-dimensional feature calculation; Geometric morphological feature calculation Curvature: Using the sliding window method (window size 5×5 pixels), according to the formula...

[0029] In the formula, Represented as curvature value; It is expressed as the radius of curvature of the curve at that point, and the curvature value of a natural shoreline is >0.15; The curvature values ​​at each point on the shoreline (R is the radius of curvature) were calculated, and the curvature values ​​of the candidate natural shoreline areas were all >0.15. Slope: Based on DEM data, the maximum gradient method is used, and the specific formula is as follows: S

[0030] In the formula, S is the slope value. , These are the elevation change rates in the x and y directions, respectively, and the slope range of the natural shoreline is 5°-45°. Vegetation cover characteristics calculation: based on the formula;

[0031] In the formula, For near-infrared reflectivity, Reflectivity in the red light band, natural coastline Value > 0.3; NDVI values ​​of the surface within a 30m buffer zone were extracted (NIR is the near-infrared band reflectance, RED is the red band reflectance). The NDVI values ​​of the candidate natural shoreline areas were all >0.3.

[0032] Texture feature calculation: Based on the GLCM gray-level co-occurrence matrix (window size 3×3, distance 1 pixel, angle 0°), the texture entropy and contrast of the shoreline region are calculated, where the entropy is calculated using the formula:

[0033] In the formula, Given the probabilities of elements i and j in the gray-level co-occurrence matrix, the entropy value of the natural shoreline is >2.0; calculate (P(i,j) is the probability of an element in the gray-level co-occurrence matrix), the entropy value of the candidate regions of the natural shoreline is >2.0; the contrast is calculated using the formula:

[0034] Step 5: Classifier training and recognition; Training set construction: The 200 manually labeled shoreline samples are associated with the geometric features (curvature, slope), vegetation cover features (NDVI value), and texture features (entropy value, contrast) calculated above to construct a feature-label dataset; Model training: In the Python environment, the Scikit-learn library was used to build a random forest classifier. The number of trees was set to 500 and the maximum depth to 10. 70% of the dataset was used as the training set and 30% as the test set to train the classification model. The model achieved a test accuracy of 92.5%. Shoreline identification: The multi-dimensional features of the study area are fused and input into a trained random forest classifier, which outputs a natural shoreline probability map with a probability value range of 0-1.0. Areas with a probability ≥ 0.7 are marked as high-confidence natural shoreline areas.

[0035] Step 6: Post-processing optimization; Noise Removal: The opening operation (erosion followed by dilation, with a structuring element of 3×3 pixel rectangle) in morphological filtering is used to remove small noise patches with an area of ​​<10m² in the natural shoreline probability map. Region merging: Using a spatial continuity constraint algorithm, separate natural shoreline segments are merged (adjacent shoreline segments with a distance of <3m are considered mergeable) to generate a complete natural shoreline outline.

[0036] Step 7: Visualize the results. Distribution map generation: In ArcGIS software, based on the shoreline type identification results, a shoreline type distribution map (1:10000) of the study area is generated, with natural shorelines marked in green and artificial shorelines marked in blue to clarify the distribution range of the two types of shorelines; Data statistics: The total length of natural shoreline in the study area was 18.6 km, accounting for 58.1% of the total river length. The length and proportion of natural shoreline in each 2 km segment were also statistically analyzed. Results Display: Displays shoreline type distribution maps, natural shoreline probability maps, and statistical reports on terminal display devices, supporting interactive operations such as layer zooming and querying.

[0037] The above implementation process was verified as follows: Accuracy verification: Thirty shoreline samples (15 natural shorelines and 15 artificial shorelines) that were not used in training were selected to verify the recognition results. The accuracy rate for natural shoreline recognition was 93.3%, the accuracy rate for artificial shoreline recognition was 90.0%, and the overall recognition accuracy rate reached 91.7%. Reasonableness verification: Based on the results of the field survey, the natural shoreline is mainly distributed in the wetland area of ​​the upper reaches of the river and the gentle slope section of the middle reaches, which is consistent with the identification results; the artificial shoreline is concentrated in the river section of the urban built-up area and the port area of ​​the lower reaches, which is consistent with the actual shoreline utilization.

[0038] Table 1 is a statistical comparison table of this method and traditional spectroscopic detection;

[0039] In summary, this embodiment strictly follows the intelligent identification method and system process for urban river natural shorelines based on multi-source remote sensing data described in this invention. Through multi-source data fusion, multi-dimensional feature extraction, and random forest classifier modeling, it achieves accurate identification of natural shorelines in the study area. The identification results are highly accurate and reasonable, providing reliable technical support for urban river ecological protection and shoreline planning and management. Furthermore, the entire implementation process is standardized and highly repeatable, and can be extended to other urban river natural shoreline identification scenarios.

[0040] Please refer to the above work process. Figures 1 to 2 .

[0041] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0042] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent identification of natural river shorelines based on multi-source remote sensing data, characterized in that, S1, Multi-source data acquisition: Acquire high-resolution remote sensing images, lidar point cloud data, and auxiliary data. The high-resolution remote sensing images are RGB and multispectral images with a resolution ≤1m. The lidar point cloud data is used to generate digital elevation models (DEM) and digital surface models (DSM). The auxiliary data includes historical land use maps and hydrogeological data. S2, Data Preprocessing: Radiometric correction, geometric registration, and water body mask extraction are performed sequentially on the acquired multi-source data; the radiometric correction is used to eliminate the influence of atmospheric scattering; the geometric registration aligns the image with the digital elevation model (DEM), and the registration error is <1 pixel; the water body mask extraction uses the normalized difference water index (NDWI) to segment the main body of the river. S3, Initial shoreline extraction: Edge detection is performed using the Canny operator to identify the water-land boundary, and then topology optimization is used to remove broken segments to generate a continuous shoreline; S4, Multi-dimensional feature calculation: Calculate geometric morphology features, vegetation cover features and texture features respectively; The geometric features include curvature and slope, where curvature is calculated using the sliding window method, and the calculation formula is as follows: In the formula, Represented as curvature value; It is expressed as the radius of curvature of the curve at that point, and the curvature value of a natural shoreline is >0.15; The slope is calculated based on the DEM using the maximum gradient method. The specific formula is as follows: S In the formula, S is the slope value. , These are the elevation change rates in the x and y directions, respectively, and the slope range of the natural shoreline is 5°-45°. S5, Classifier Training and Recognition: Construct a training set containing natural and artificial shoreline samples. The geometric morphological features, vegetation cover features, and texture features calculated in S4 are fused and input into a random forest classifier. The output is a natural shoreline probability map with a probability value range of 0-1.

0. The random forest classifier has 500 trees and a maximum depth of 10. S6, Post-processing optimization: Morphological filtering is used to remove small noise patches, and adjacent natural shoreline segments are merged through spatial continuity constraints; S7, Results Visualization Output: Generate a shoreline type distribution map, clarify the distribution range of natural and artificial shorelines, and statistically analyze the proportion and segment length of natural shorelines.

2. The intelligent identification method for natural river shorelines based on multi-source remote sensing data according to claim 1, characterized in that, The high-resolution remote sensing imagery mentioned in S1 includes WorldView-3 imagery, and the lidar point cloud data includes airborne LiDAR point cloud data.

3. The intelligent identification method for natural river shorelines based on multi-source remote sensing data according to claim 1, characterized in that, The width of the shoreline buffer zone mentioned in S1 is set to 20-30m. The vegetation cover feature is extracted by NDVI enhancement to obtain the vegetation index within the shoreline buffer zone. The calculation formula is as follows: In the formula, For near-infrared reflectivity, Reflectivity in the red light band, natural coastline Value > 0.3; The texture features are calculated based on the GLCM gray-level co-occurrence matrix, using entropy and contrast values. The entropy calculation formula is as follows: In the formula, Let i be the probability of elements i and j in the gray-level co-occurrence matrix, and let the entropy of the natural shoreline be > 2.

0. The formula for calculating contrast is: 。 4. The intelligent identification method for natural river shorelines based on multi-source remote sensing data according to claim 1, characterized in that, The training set described in S1 was obtained by manually annotating urban river shoreline samples, which covered shoreline areas with different landform types, vegetation cover, and artificial modification intensity.

5. A river natural shoreline intelligent identification system based on multi-source remote sensing data, adapted to the river natural shoreline intelligent identification method based on multi-source remote sensing data as described in any one of claims 1-4, characterized in that, It includes hardware components and software modules. The hardware components include a remote sensing data server, a GPU computing node, and a terminal display device. The software modules include a data acquisition module, a data preprocessing module, a shoreline initial extraction module, a multi-dimensional feature calculation module, a classification and recognition module, a post-processing optimization module, and a result output module. The data acquisition module is used to collect high-resolution remote sensing images, lidar point cloud data, and auxiliary data. The high-resolution remote sensing images are RGB / multispectral images with a resolution ≤1m. The lidar point cloud data is used to generate DEM and DSM. The auxiliary data includes historical land use maps and hydrogeological data. The data preprocessing module is used to perform radiometric correction, geometric registration, and water body mask extraction on the acquired multi-source data. The radiometric correction eliminates the influence of atmospheric scattering, the geometric registration aligns the image with the DEM with an error of <1 pixel, and the water body mask extraction uses NDWI exponential segmentation of the river body. The initial shoreline extraction module uses the Canny operator to perform edge detection and identify the water-land boundary, and generates a continuous shoreline through topology optimization; The multi-dimensional feature calculation module is used to calculate geometric features, vegetation cover features, and texture features. The geometric features include curvature and slope. Curvature is calculated using the sliding window method, and slope is calculated based on DEM. The vegetation cover feature is the NDVI value within the shoreline buffer. The texture feature is the entropy value of the GLCM gray-level co-occurrence matrix. The classification and recognition module includes a training set construction unit and a random forest classifier. The training set construction unit is used to store manually labeled natural / artificial shoreline samples. The random forest classifier receives multi-dimensional feature fusion data and outputs a natural shoreline probability map. The random forest classifier has 500 trees and a maximum depth of 10. The post-processing optimization module uses morphological filtering to remove noise and spatial continuity constraints to merge adjacent natural shoreline segments. The result output module is used to generate a shoreline type distribution map and statistically analyze the proportion and segment length of natural shorelines, and display it on the terminal display device.

6. The intelligent river shoreline identification system based on multi-source remote sensing data according to claim 5, characterized in that, The remote sensing data server is used to store high-resolution remote sensing images and lidar point cloud data acquired by satellites and drones, and the GPU computing node is used to accelerate the multi-dimensional feature extraction and random forest classifier operation.

7. The intelligent river shoreline identification system based on multi-source remote sensing data according to claim 5, characterized in that, In the multi-dimensional feature calculation module, the curvature threshold of the natural shoreline is >0.15, the slope range is 5°-45°, the NDVI threshold is >0.3, and the entropy threshold is >2.

0.

8. The intelligent river shoreline identification system based on multi-source remote sensing data according to claim 5, characterized in that, The width of the shoreline buffer zone is set to 30m, and the morphological filtering uses a combination of opening and closing operations to remove small noise patches.

Images