A method and system for remote sensing identification of spillways with multi-feature support

Abstract

The application discloses a kind of multi-feature supported spillway remote sensing identification method and system, and the object level target is obtained by image segmentation algorithm;Spectrum, shape, geometric stereo feature of object level target is extracted, and preliminary extraction of spillway object is realized by multi-feature joint decision;The spatial direction and distance relationship between objects are comprehensively considered, and the fine spot around the preliminary identification result is identified and integrated again, contour modification is carried out simultaneously, and finally the relatively neat spillway boundary vector is output.This method specifically designs the geometric stereo feature parameters of spillway to highlight the unique features of spillway in three-dimensional space, effectively overcoming the difficulty of distinguishing spillway from general buildings.At the same time, the method also provides an automatic remote sensing spillway identification process to improve the accuracy of spillway identification, which has important practical application value in the fields of spillway remote sensing monitoring and water conservancy construction.

Description

Technical Field

[0001] The method of this invention belongs to the field of remote sensing monitoring and water conservancy informatization, and specifically relates to a method and system for remote sensing identification of spillways supported by multiple features. Background Technology

[0002] Remote sensing technology, as an important component of smart water conservancy, is based on remote sensing imagery and establishes a high-precision and highly automated remote sensing information extraction technology. This technology can efficiently identify, monitor, and update data of basic spatial elements in water conservancy, and has significant value in scientific research and engineering applications.

[0003] Spillways, as key facilities for ensuring the safe operation of water conservancy projects such as reservoirs, play a vital role in flood control. Utilizing remote sensing to automatically identify spillways enables regular monitoring and data updates at a relatively low cost, providing information support for detecting anomalies such as spillway blockages and occupancy, thereby ensuring the safe operation of dams and reservoirs.

[0004] Currently, there is a gap in research on spillway identification based on remote sensing imagery both domestically and internationally. The main challenges are: ① Spillways are typically small in area, occupying only a tiny region in large-scale remote sensing images, and may even appear as just a few pixels or a single pixel in low-to-medium resolution images; ② Spillways are often built on one side of dams, surrounded by other hydraulic structures such as dams and canals, creating a complex background that makes accurate extraction difficult; ③ Spillways lack unique spectral and shape features, making them difficult to distinguish from surrounding regular structures such as roads, bridges, and buildings. Therefore, current research on extracting basic spatial elements of hydraulic infrastructure using remote sensing imagery mostly focuses on watershed natural elements such as rivers, lakes, and reservoirs, with relatively little research on extracting basic spatial elements of hydraulic infrastructure such as spillways, dams, and canals. Summary of the Invention

[0005] To address the aforementioned technical challenges, this invention proposes a multi-feature-supported remote sensing identification method for spillways. Targeting the difficulties in spillway identification, this method fully utilizes domestically produced high-resolution satellite data. Based on a systematic analysis of the spectral and geometric features of spillways in remote sensing imagery, a targeted geometric three-dimensional feature index for spillways is designed to highlight their unique characteristics in three-dimensional space. Furthermore, a multi-feature-supported automatic spillway identification method is proposed, aiming to achieve intelligent spillway identification based on high-resolution remote sensing imagery and provide partial technical support for the remote sensing monitoring technology system of basic water conservancy spatial elements.

[0006] The technical solution of this invention is a multi-feature supported remote sensing identification method for spillways, the specific steps of which are as follows:

[0007] Step 1: Acquire high-resolution raw remote sensing satellite imagery, obtain a preprocessed first image through image preprocessing, segment the first image, and obtain object-level targets.

[0008] Step 2. Based on object-oriented thinking, extract multi-dimensional features at the object level, and achieve preliminary identification of the spillway through joint constraints of multi-dimensional features;

[0009] Step 3. Perform post-processing operations on the preliminary identification results. Traverse the surrounding fragmented patches that have been identified as spillway objects, consider the spatial context features of the spillway, and identify and integrate the fragmented patches by combining features such as distance relationship and spatial direction. At the same time, perform contour shaping to obtain a regular spillway boundary.

[0010] Furthermore, obtaining the object-level target in step 1 is divided into two parts: preprocessing and image segmentation, and the specific implementation method is as follows:

[0011] Step 1.1 Preprocessing: Perform orthorectification, radiometric calibration, atmospheric correction, image fusion, and image cropping on the original image to obtain the first image after preprocessing;

[0012] Step 1.2 Image Segmentation: The Mean-Shift algorithm is used to segment the preprocessed image to obtain object-level targets. Further, the multidimensional features in Step 2 include spectral features, shape features, and geometric solid features.

[0013] Furthermore, step 2, which involves removing water bodies and vegetation from the target-level objectives based on spectral features, specifically involves:

[0014] The Normalized Difference Water Index (NDWI) and Normalized Difference Vegetation Index (NDVI) of the calculated objects were used. The Otsu's method (OTSU) was employed to calculate the screening threshold, excluding water bodies and vegetation objects. The formula is as follows:

[0015]

[0016]

[0017] In the formula: ρ represents the band reflectance, and Green, Red, and NIR represent the green, red, and near-infrared bands of the remote sensing image, respectively.

[0018] Furthermore, in step 2, the irregular interference objects in the object-level target are removed based on shape features. Specifically, the rectangle fitting degree of each object in the object-level target is calculated, and a rectangle fitting degree threshold is preset. Interference objects that are less than the preset rectangle fitting degree threshold are filtered out.

[0019] The formula for calculating the rectangle fit (Rect) is as follows:

[0020]

[0021] In the formula: S0 is the area of ​​the object, S Rect Rect represents the area of ​​the bounding rectangle of an object, reflecting the degree to which the object fills its bounding rectangle.

[0022] Furthermore, step 2, which involves removing other types of regular rectangular artificial structures from the object-level target based on geometric solid features, specifically involves:

[0023] For each object element, establish a minimum bounding rectangle, and define the direction of the longer side of the minimum bounding rectangle as its primary direction, and the direction perpendicular to the longer side as its secondary direction. Draw equidistant straight lines parallel to the primary direction. There are n straight lines that pass through the target object. The intersection point of the i-th straight line and the target object when the i-th straight line first enters the target object is P. i The point where the straight line intersects the target object for the last time is P. i ′, then P i With P i The absolute value of the elevation difference at point ′ is set as the three-dimensional feature on the i-th line along the main direction of the object;

[0024] Draw m equally spaced straight lines parallel to the secondary direction, passing through the target object. The intersection point of the j-th line and the target object when the j-th line first enters the target object is S. j The point where the straight line intersects the target object for the last time is S. j ′, then S j With S j The absolute value of the elevation difference at point ′ is set as the solid feature on the j-th line of the object's secondary direction;

[0025] The Geometric Solid Index (SGSI) is used to distinguish spillways from other regular rectangular man-made structures.

[0026]

[0027] ΔP i (>30) and ΔS j (<10) It can distinguish spillways from other types of regular rectangular artificial structures.

[0028] Furthermore, step 3 includes the following sub-steps:

[0029] Step 3.1 Combining spatial context features with fragmented map traversal, the surrounding fragmented map patches identified as spillway objects are traversed. If there are map patches that are also identified as spillways, they are directly integrated with the identification results. The remaining unidentified fragmented map patches meet the following conditions: ① they exist within a 5m range of the main direction of the spillway object; ② the angular deviation between the main direction of their circumscribed rectangle and the main direction of the spillway object does not exceed 20°; ③ their rectangle fitting degree is >0.6. Then they are determined to be spillway-related map patches and integrated with the identified results.

[0030] Step 3.2 Contour shaping: Establish the minimum bounding rectangle of the recognition result and output the regularized boundary vector of the spillway.

[0031] On the other hand, the present invention provides a multi-feature-supported spillway remote sensing identification system, comprising:

[0032] The object-level target acquisition module is used to acquire high-resolution raw remote sensing satellite images, obtain a preprocessed first image through image preprocessing, perform image segmentation on the first image, and acquire object-level targets.

[0033] The preliminary identification module is used to extract multi-dimensional features at the object level based on object-oriented thinking, and to achieve preliminary identification of the spillway through joint constraints of multi-dimensional features;

[0034] The integration module is used to perform post-processing operations on the preliminary identification results. It traverses the surrounding fragmented patches that have been identified as spillway objects, considers the spatial context features of the spillway, and identifies and integrates the fragmented patches by combining features such as distance relationship and spatial direction. At the same time, it performs contour shaping to obtain a regular spillway boundary.

[0035] The multi-feature-supported spillway remote sensing identification system performs the steps in the multi-feature-supported spillway remote sensing identification method.

[0036] Compared with the prior art, the present invention has the following features and beneficial effects:

[0037] This invention addresses the technical challenge of accurately identifying small targets in spillways based on remote sensing imagery during water conservancy informatization construction. It proposes a multi-feature-supported remote sensing identification method for spillways. This invention employs an object-oriented approach, fully exploring the object-level features of spillways in remote sensing imagery. Through mathematical abstraction of multi-dimensional features, a complex feature model of the spillway is constructed, enabling effective differentiation between the spillway and general land features, and even interfering features such as man-made structures. The invention performs post-processing on the identification results, considering the spatial context features of the spillway object to achieve secondary identification and integration of surrounding fragmented patches, thereby maximizing the integrity of the spillway extraction. Furthermore, this invention designs an automated spillway remote sensing identification process, improving the efficiency and accuracy of spillway remote sensing monitoring. It provides partial technical support for the remote sensing monitoring technology system of basic water conservancy spatial elements and has significant practical application value in scenarios such as water conservancy informatization construction. Attached Figure Description

[0038] Figure 1 A schematic diagram illustrating the object-level target acquisition for the Mean-shift segmentation algorithm;

[0039] Figure 2 A schematic diagram illustrating spectral feature filtering using OTSU adaptive threshold selection.

[0040] Figure 3 This is a schematic diagram illustrating the SGSI extraction principle.

[0041] Figure 4 Schematic diagram of post-processing for spillway object identification;

[0042] Figure 5 This is a schematic diagram of the output results for automated recognition. Detailed Implementation

[0043] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the invention.

[0044] Example 1

[0045] The example data for this invention selects spillways around two dams: Hekou Village Reservoir in Jiyuan City, Henan Province, and Nuozhadu Reservoir (hydropower station) on the lower reaches of the Lancang River. Python is chosen as the programming language, and ArcPy and GDAL are used as geoprocessing algorithm libraries. ArcPy is a Python toolkit provided by ArcGIS, allowing users to use Python to call various toolboxes provided by ArcGIS for batch processing. GDAL is an open-source GIS algorithm library that can be combined with important Python packages such as NumPy to achieve image data processing.

[0046] The specific implementation process steps are as follows:

[0047] Step 1: Based on actual business needs, select GF-7 remote sensing images corresponding to the spillways of three dams—Hekou Village Reservoir in Jiyuan City, Henan Province; Nuozhadu Reservoir (hydropower station) in the lower reaches of the Lancang River; and Longyangxia Reservoir in the upper reaches of the Yellow River—as data sources. The GF-7 images include GF-7 back-view multispectral images, GF-7 back-view panchromatic images, and GF-7 front-view panchromatic images. The panchromatic images have a resolution better than 0.8m, and the multispectral images have a resolution of 3.2m.

[0048] Step 1.1 Preprocessing. Using professional software such as ENVI and ArcGIS, orthorectification, radiometric calibration, atmospheric correction, image fusion, and image cropping are performed on the GF-7 remote sensing images to obtain preprocessed satellite fused images. These fused images retain the advantages of GF-7's high spatial resolution and multispectral properties, effectively improving the accuracy of small target identification.

[0049] Step 1.2 Image Segmentation. The Mean-Shift algorithm is used for density center estimation in multivariate feature spaces. It iterative calculations based on gradient climbing of probability density are used to find local optima, making it an effective iterative clustering algorithm. In this embodiment, the Mean-Shift algorithm is used to segment the GF-7 fused image. It requires no prior knowledge, automatically determines the number of categories based on image characteristics, and has good noise resistance, effectively preserving region boundary information. The segmentation results are converted into vectors to obtain object-level targets, such as... Figure 1 As shown, (a) is a schematic diagram of the target at the Nuozhadu Hydropower Station, and (b) is a schematic diagram of the target at the Hekou Village Reservoir. This step can be automated by using Python to call Arcpy. In this embodiment, the Mean-Shift algorithm is used to segment the preprocessed image, and the specific principle is shown in formula (1):

[0050]

[0051] In the formula: x idenoted by , h represents the region radius, g(·) represents the kernel function, and w(·) represents the introduced weights.

[0052] Step 2. Read the object vector file generated in Step 1.2, obtain the band information of each target object, calculate the multi-dimensional features such as the object's spectral features, shape features, and geometric solid features, and then perform joint constraints on the multi-dimensional features to achieve preliminary identification of the spillway. All operations in Step 2 can be automated using Python programming.

[0053] Step 2.1 Spectral Characteristics. Obtain the band information of the object and calculate the average NDWI of the pixels contained in the object. The value range of NDWI is [-1, 1]. The larger the value, the greater the probability that the object is a water body. The formulas for calculating the Normalized Difference Water Index (NDWI) and Normalized Difference Vegetation Index (NDVI) are shown in equations (2) and (3):

[0054]

[0055]

[0056] OTSU can divide an image into two parts, background and target, based on its grayscale characteristics, and then select a threshold that maximizes the inter-class variance between the two parts (as shown in the attached figure). Figure 2 As shown in the figure, (a) taking Nuozhadu as an example, OTSU adaptively selects an NDVI threshold of 0.199, and the part greater than 0.199 is regarded as vegetation removal; (b) taking Nuozhadu as an example, OTSU adaptively selects an NDWI threshold of 0.00, and the part greater than 0.00 is regarded as water body removal.

[0057] By using OTSU to adaptively calculate the NDWI threshold, different environmental characteristics can be adapted to accurately filter and remove water bodies from images. Similarly, by calculating the average NDVI of the pixels contained in an object and using OTSU to adaptively select a threshold, vegetation objects can be removed.

[0058] Step 2.2 Shape Features. Obtain the area information of the object. At the same time, create the minimum bounding rectangle of the object and obtain the area information of its bounding rectangle. Calculate the rectangle fit of the object and remove objects with a rectangle fit of less than 0.6. This can filter out irregular interference objects such as excavations, slopes, and bare land, and retain objects with relatively regular shapes.

[0059] The specific rectangle fitting degree is shown in equation (4).

[0060]

[0061] In the formula: S0 is the area of ​​the object, S Rect Rect represents the area of ​​the bounding rectangle of an object, reflecting the degree to which the object fills its bounding rectangle.

[0062] Step 2.3 Geometric Features of the Spillway. The GF-7 satellite has forward and backward stereo imaging capabilities, which can produce high-precision DSMs to provide elevation information for the study area. Matching the DSM with the fused imagery allows for the extraction of SGSIs, as illustrated in the attached diagram. Figure 3 As shown, (1) Determine the orientation of the object. First, establish the minimum bounding rectangle of the object. Take the direction of the long side of the minimum bounding rectangle as the main direction of the object, and the direction perpendicular to the long side as the secondary direction of the object; (2) Extract the three-dimensional features of the main direction. Draw n lines parallel to the main direction and passing through the target object. The intersection point of the line with the target object when it first enters the target object is P. i The point where the straight line intersects the target object for the last time is P. i ′, where i is the index of the line, obtain P i With P i The elevation of point ′ is calculated and the difference is obtained. The absolute value of the elevation difference is the three-dimensional feature of the i-th line in the main direction. (3) Extraction of three-dimensional features in the secondary direction: draw m lines parallel to the secondary direction and passing through the target object. The intersection point of the line with the target object when it first enters the target object is S. j The point where the straight line intersects the target object for the last time is S. j ', where j is the index of the line. Obtain S j With S j The elevation of point ′ is calculated and the difference is obtained. The absolute value of the elevation difference is the three-dimensional feature of the j-th line in the secondary direction. SGSI cleverly highlights the difference between the spillway and other buildings in terms of geometric three-dimensional features. The SGSI calculation formula is shown in Equation (5).

[0063]

[0064] As shown in Table 1, using ΔP i (>30) and ΔS j (<10) It can distinguish spillways from other types of regular rectangular artificial structures.

[0065] Table 1 Comparison of geometric features of different types of land cover

[0066]

[0067] Spectral features can be used to remove water bodies and vegetation, while shape features can be used to remove bare soil, excavated land, or irregular buildings, thus retaining relatively regular man-made structures. Geometric features can effectively distinguish spillways from other regular rectangular man-made structures, thus effectively extracting spillways.

[0068] Step 3. The accuracy of object-oriented target recognition is affected by the image segmentation accuracy. In some cases, image segmentation may fail to accurately capture the complete boundaries of objects, and target features may be over-segmented into fragmented objects. Small, fragmented spillway patches are easily missed due to their indistinct geometric features. To ensure the integrity of the spillway, the spatial context features of the spillway are considered. The surrounding fragmented patches of the already identified spillway objects are traversed, and features such as distance relationships and spatial orientation are combined to identify and integrate the fragmented patches. At the same time, contour shaping is performed to obtain a regular spillway boundary. All operations in Step 3 can be automated using Python programming.

[0069] Step 3.1 Traversal of fragmented map features combined with spatial context. The surrounding fragmented map features identified as spillway objects are traversed. If a map feature also identified as a spillway exists, it is directly integrated with the identification results. Other unidentified fragmented map features that meet the following conditions are considered spillway-related: ① they exist in the vicinity of the spillway object's main direction; ② the main direction of their circumscribed rectangle is basically consistent with the main direction of the spillway object; ③ their rectangle fitting degree is >0.6. These are then identified as spillway-related map features and integrated with the already identified results, as shown in the attached figure. Figure 4 .

[0070] Step 3.2 Contour Shaping. Establish the minimum bounding rectangle of the recognition results and output the regularized boundary vector of the spillway, as shown in the attached figure. Figure 5 (a) is a schematic diagram of the identification output results of Nuozhadu Hydropower Station, and (b) is a schematic diagram of the identification output results of Hekou Village Reservoir.

[0071] Example 2

[0072] This embodiment presents a multi-feature-supported spillway remote sensing identification system, comprising:

[0073] The object-level target acquisition module is used to acquire high-resolution raw remote sensing satellite images, obtain a preprocessed first image through image preprocessing, perform image segmentation on the first image, and acquire object-level targets.

[0074] The preliminary identification module is used to extract multi-dimensional features at the object level based on object-oriented thinking, and to achieve preliminary identification of the spillway through joint constraints of multi-dimensional features;

[0075] The integration module is used to perform post-processing operations on the preliminary identification results. It traverses the surrounding fragmented patches that have been identified as spillway objects, considers the spatial context features of the spillway, and identifies and integrates the fragmented patches by combining features such as distance relationship and spatial direction. At the same time, it performs contour shaping to obtain a regular spillway boundary.

[0076] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0077] Furthermore, those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, any of the claimed embodiments can be used in any combination.

Claims

1. A method for flood spillway remote sensing identification with multi-feature support, characterized in that, Includes the following steps: Step 1: Acquire high-resolution raw remote sensing satellite imagery, obtain a preprocessed first image through image preprocessing, segment the first image, and obtain object-level targets. Step 2. Based on object-oriented principles, extract object-level multidimensional features and achieve preliminary identification of the spillway through joint constraints of multidimensional features; the multidimensional features include spectral features, shape features, and geometric solid features; Based on geometric solid features, other types of regular rectangular artificial buildings are removed from the object-level target, specifically: For each object element, establish a minimum bounding rectangle, and define the direction of the longer side of the minimum bounding rectangle as its primary direction, and the direction perpendicular to the longer side as its secondary direction. Draw equidistant straight lines parallel to the primary direction; there exist n straight lines passing through the target object. The point where the straight line intersects the target object the first time it enters the target object is: The point where the straight line intersects the target object for the last time is... Then and The absolute value of the elevation difference of a point is set as the first point along the principal direction of the object. Three-dimensional features along the lines; Parallel to the sub-direction, there are m straight lines passing through the target object, the first straight line entering the target object is , the intersection point of the straight line and the target object is , the intersection point of the last straight line passing out of the target object and the target object is , then the absolute value of the elevation difference between and is set as the three-dimensional feature on the th line in the sub-direction of the object. Constructing geometric stereoscopic indices for spillways Differentiate spillways from other similar regular rectangular artificial structures, then and Distinguishing spillways from other regular rectangular artificial structures; Step 3. Perform post-processing operations on the preliminary identification results. Traverse the surrounding fragmented patches that have been identified as spillway objects, consider the spatial context features of the spillway, and identify and integrate the fragmented patches by combining distance relationships and spatial direction features. At the same time, perform contour shaping to obtain a regular spillway boundary.

2. The method as claimed in claim 1, wherein the method is a multi-feature supported spillway identification method. Step 1, obtaining the object-level target, is divided into two parts: preprocessing and image segmentation. The specific implementation method is as follows: Step 1.1 Preprocessing: Perform orthorectification, radiometric calibration, atmospheric correction, image fusion, and image cropping on the original image to obtain the first image after preprocessing; Step 1.2 Image Segmentation: The Mean-Shift algorithm is used to segment the preprocessed image to obtain object-level targets.

3. The method as claimed in claim 1, wherein the method is a multi-feature supported spillway identification method. Step 2, which involves removing water bodies and vegetation from the target-level objectives based on spectral features, specifically involves: The Normalized Difference Water Index (NDWI) and Normalized Difference Vegetation Index (NDVI) of the calculated objects were used. The Otsu's method (OTSU) was employed to calculate the screening threshold, excluding water bodies and vegetation objects. The formula is as follows: In the formula: , , respectively represent the green, red, near-infrared band reflectance of the remote sensing image.

4. The method as claimed in claim 1, wherein the method is a multi-feature supported spillway identification method. In step 2, irregular interference objects in the object-level target are removed based on shape features. Specifically, the rectangle fitting degree of each object in the object-level target is calculated, and a rectangle fitting degree threshold is preset. Interference objects that are less than the preset rectangle fitting degree threshold are filtered out. Rectangle fit The calculation formula is as follows: In the formula: For the area of ​​the object, The area of ​​the bounding rectangle of the object. It reflects the degree to which an object fills its bounding rectangle.

5. The method for remote sensing identification of spillways supported by multiple features according to claim 1, characterized in that: Step 3 includes the following sub-steps: Step 3.1 Combining spatial context features with fragmented map traversal, the surrounding fragmented map patches identified as spillway objects are traversed. If there are map patches that are also identified as spillways, they are directly integrated with the identification results. The remaining unidentified fragmented map patches meet the following conditions: ① they exist within a 5m range of the main direction of the spillway object; ② the angular deviation between the main direction of their circumscribed rectangle and the main direction of the spillway object does not exceed 20°; ③ their rectangle fitting degree is >0.

6. Then they are determined to be spillway-related map patches and integrated with the identified results. Step 3.2 Contour shaping: Establish the minimum bounding rectangle of the recognition result and output the regularized boundary vector of the spillway.

6. A multi-feature-supported remote sensing identification system for spillways, characterized in that, include: The object-level target acquisition module is used to acquire high-resolution raw remote sensing satellite images, obtain a preprocessed first image through image preprocessing, perform image segmentation on the first image, and acquire object-level targets. The preliminary identification module is used to extract multi-dimensional features at the object level based on object-oriented thinking, and to achieve preliminary identification of the spillway through joint constraints of multi-dimensional features; The integration module is used to perform post-processing operations on the preliminary identification results. It traverses the surrounding fragmented patches that have been identified as spillway objects, considers the spatial context features of the spillway, and identifies and integrates the fragmented patches by combining distance relationships and spatial direction features. At the same time, it performs contour shaping to obtain a regular spillway boundary. The multi-feature-supported spillway remote sensing identification system performs the steps of the multi-feature-supported spillway remote sensing identification method as described in any one of claims 1-5.

Images