Seafloor terrain processing method based on terrain break zone and object-level geomorphic attribute

Abstract

This invention discloses a seabed topography processing method based on topographic abrupt change zones and object-level geomorphic attributes, belonging to the field of image processing technology. It is used for seabed topography processing, including: determining the monitoring area; acquiring and preprocessing the seabed DEM of the monitoring area; calculating seabed topographic factors; constructing a topographic abrupt change index map based on the seabed topographic factors and extracting topographic abrupt change zones; generating candidate geomorphic objects based on the topographic abrupt change zones; extracting object-level geomorphic attributes based on the spatial extent of the candidate geomorphic objects; generating an object-numbered raster map according to the raster numbers of the candidate geomorphic objects; and converting the boundary of each candidate geomorphic object into a vector boundary with geographic coordinate attributes to form a vectorized segmentation result. This invention, by constructing a topographic abrupt change index and extracting object-level geomorphic attributes, achieves stable, interpretable, and training-sample-free topographic unit segmentation and spatial representation of seabed digital elevation models.

Description

Technical Field

[0001] This invention discloses a method for processing seabed topography based on topographic abrupt change zones and object-level geomorphic attributes, belonging to the field of image processing technology. Background Technology

[0002] In seafloor topography analysis, geomorphic unit segmentation and attribute extraction based on Digital Elevation Models (DEMs) are fundamental tasks in marine geological surveys and mapping. Existing methods often employ single topographic factors such as slope and curvature, combined with threshold segmentation or watershed algorithms, to directly process the DEM. However, due to the complexity of seafloor topography and the fuzzy boundaries of geomorphic features, segmentation results are often unstable, leading to oversegmentation or undersegmentation, making it difficult to form continuous objects corresponding to real geomorphic units. Furthermore, traditional methods typically only output raster classification results or simple statistics, lacking systematic extraction and vectorized spatial representation of object-level geomorphic attributes (such as center-edge elevation difference, compactness, aspect ratio, etc.), limiting subsequent analytical applications. Therefore, there is an urgent need for a processing method that can stably segment seafloor topographic units, enrich object-level attribute descriptions, and directly output spatialized results. Summary of the Invention

[0003] The purpose of this invention is to provide a method for processing seabed topography based on topographic abrupt change zones and object-level geomorphic attributes, in order to solve the problems of unstable candidate object segmentation, lack of object-level geomorphic attribute system extraction, and insufficient spatialization expression in the prior art.

[0004] Seafloor topography processing methods based on terrain abrupt change zones and object-level geomorphic attributes include: S1. Determine the monitoring area, acquire the seabed DEM of the monitoring area and preprocess it, and calculate the seabed topographic factor based on the preprocessing results. S2. Construct a topographic abrupt change index map based on seafloor topographic factors, and extract topographic abrupt change zones using the threshold method; use the topographic abrupt change zones as object boundary constraints, and generate candidate landform objects using connected component analysis, distance transformation, and watershed segmentation methods; extract object-level landform attributes based on the spatial extent of the candidate landform objects. S3. Generate an object number raster map based on the raster number of the candidate landform objects, and convert the boundary of each candidate landform object into a vector boundary with geographic coordinate attributes to form a vectorized segmentation result.

[0005] S1 includes preprocessing such as outlier removal, invalid value imputation, spatial pruning, resampling, and normalization; the seafloor topographic factors include at least the elevation gradient. ,slope Slope change rate and curvature ; based on Construct a slope map based on Construct a curvature graph.

[0006] S2 includes S2.1, the terrain abrupt change index. Depend on , and In summary: ; ; In the formula, For normalization function, , , These are the weighting coefficients.

[0007] S2 includes, S2.2, using quantile thresholding or adaptive thresholding methods to... Areas exceeding a threshold are defined as terrain abrupt change zones. Regions less than or equal to the threshold are defined as non-mutation regions, and these non-mutation regions are considered as internal regions of candidate landform objects.

[0008] S2 includes S2.3, and the process of generating candidate landform objects includes: Using abrupt terrain change zones as object boundary constraints, we perform connected component analysis on non-abrupt regions to extract all connected pixel components, and use these connected pixel components as regions to be segmented. For each region to be segmented, perform a distance transformation, calculate the Euclidean distance from each pixel to the nearest terrain abrupt change zone, and obtain the internal distance map of the region; Local maxima points are identified based on distance maps and used as segmentation markers in the watershed segmentation method; The watershed segmentation method is used to divide the current area to be segmented into several candidate landform objects; Calculate the area of ​​each candidate landform object. After segmenting the candidate landform objects, set a minimum object size threshold. Candidate landform objects with fewer than the minimum object size threshold are treated as noise objects and removed. Candidate landform objects with more than or equal to the minimum object size threshold are retained as valid candidate landform objects. For valid candidate landform objects, sort them from largest to smallest area, and renumber them using a continuous integer coding method based on the sorting results.

[0009] S2 includes S2.4, which extracts object-level geomorphic attributes for each candidate geomorphic object in the candidate geomorphic object set based on the spatial range of the candidate geomorphic object in the seabed DEM, slope map, curvature map and topographic abrupt change index map.

[0010] S3 includes the vectorized segmentation results, which include candidate geomorphic object numbers, geometric boundaries, and object-level geomorphic attributes. The output results are saved in GIS vector data format.

[0011] Object-level geomorphic attributes include at least the center-edge elevation difference, aspect ratio, compactness, and topographic abrupt change index.

[0012] Compared to existing technologies, this invention offers the following advantages: Firstly, by constructing a terrain abrupt change index that integrates elevation gradient, slope change rate, and curvature, and using its high-value areas as hard boundary constraints for watershed segmentation, this invention effectively suppresses over-segmentation and under-segmentation, significantly improving the spatial correspondence stability between candidate landform objects and real terrain units. Secondly, the system extracts object-level landform attributes, including center-edge elevation difference, aspect ratio, compactness, and terrain abrupt change index, providing rich features for positive and negative terrain discrimination, morphological quantification, and subsequent analysis. Finally, it directly generates object-numbered rasters and attribute-bearing vector boundaries, achieving spatialized representation of segmentation results and seamless integration with GIS, facilitating seabed landform mapping, statistical analysis, and geological interpretation. The entire method requires no manual annotation or training samples, is highly automated, and is applicable to various types of seabed DEM data. Attached Figure Description

[0013] Figure 1 This is a flowchart of the technology of this invention; Figure 2 This is a seabed DEM according to an embodiment of the present invention; Figure 3 This refers to the terrain abrupt change index in an embodiment of the present invention; Figure 4 This is a terrain abrupt change zone in an embodiment of the present invention; Figure 5 These are candidate landform objects in embodiments of the present invention; Figure 6 This is the classification grid result of an embodiment of the present invention; Figure 7 This is the result of vector boundary superposition in an embodiment of the present invention; Figure 8 This is the classification result of the seabed topography unit 0001 in the embodiment of the present invention; Figure 9 This is the classification result of the seabed topographic unit 0039 in the embodiment of the present invention; Figure 10 This is the classification result of the seabed topography unit 0030 in the embodiment of the present invention; Figure 11 This is the classification result of the seabed topography unit 0061 in the embodiment of the present invention; Figure 12 This is the classification result of the seabed topography unit 0083 in the embodiment of the present invention; Figure 13This is the classification result of the seabed topographic unit 0091 of the embodiment of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention are described clearly and completely below. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0015] Seafloor topography processing methods based on terrain abrupt change zones and object-level geomorphic attributes include: S1. Determine the monitoring area, acquire the seabed DEM of the monitoring area and preprocess it, and calculate the seabed topographic factor based on the preprocessing results. S2. Construct a topographic abrupt change index map based on seafloor topographic factors, and extract topographic abrupt change zones using the threshold method; use the topographic abrupt change zones as object boundary constraints, and generate candidate landform objects using connected component analysis, distance transformation, and watershed segmentation methods; extract object-level landform attributes based on the spatial extent of the candidate landform objects. S3. Generate an object number raster map based on the raster number of the candidate landform objects, and convert the boundary of each candidate landform object into a vector boundary with geographic coordinate attributes to form a vectorized segmentation result.

[0016] S1 includes preprocessing such as outlier removal, invalid value imputation, spatial pruning, resampling, and normalization; the seafloor topographic factors include at least the elevation gradient. ,slope Slope change rate and curvature ; based on Construct a slope map based on Construct a curvature graph.

[0017] S2 includes S2.1, the terrain abrupt change index. Depend on , and In summary: ; ; In the formula, For normalization function, , , These are the weighting coefficients.

[0018] S2 includes, S2.2, using quantile thresholding or adaptive thresholding methods to... Areas exceeding a threshold are defined as terrain abrupt change zones. Regions less than or equal to the threshold are defined as non-mutation regions, and these non-mutation regions are considered as internal regions of candidate landform objects.

[0019] S2 includes S2.3, and the process of generating candidate landform objects includes: Using abrupt terrain change zones as object boundary constraints, we perform connected component analysis on non-abrupt regions to extract all connected pixel components, and use these connected pixel components as regions to be segmented. For each region to be segmented, perform a distance transformation, calculate the Euclidean distance from each pixel to the nearest terrain abrupt change zone, and obtain the internal distance map of the region; Local maxima points are identified based on distance maps and used as segmentation markers in the watershed segmentation method; The watershed segmentation method is used to divide the current area to be segmented into several candidate landform objects; Calculate the area of ​​each candidate landform object. After segmenting the candidate landform objects, set a minimum object size threshold. Candidate landform objects with fewer than the minimum object size threshold are treated as noise objects and removed. Candidate landform objects with more than or equal to the minimum object size threshold are retained as valid candidate landform objects. For valid candidate landform objects, sort them from largest to smallest area, and renumber them using a continuous integer coding method based on the sorting results.

[0020] S2 includes S2.4, which extracts object-level geomorphic attributes for each candidate geomorphic object in the candidate geomorphic object set based on the spatial range of the candidate geomorphic object in the seabed DEM, slope map, curvature map and topographic abrupt change index map.

[0021] S3 includes the vectorized segmentation results, which include candidate geomorphic object numbers, geometric boundaries, and object-level geomorphic attributes. The output results are saved in GIS vector data format.

[0022] Object-level geomorphic attributes include at least the center-edge elevation difference, aspect ratio, compactness, and topographic abrupt change index.

[0023] The process of calculating seabed topographic factors in this invention includes, assuming the seabed DEM is... First gradient in the horizontal direction and the first gradient in the vertical direction They are respectively: ; ; elevation gradient for: ; In the formula, for The first-order gradient in the horizontal direction, for First-order gradient in the vertical direction; based on Calculate slope : ; right Calculate the first-order gradient and the rate of change of slope. : ; In the formula, for The first-order gradient in the horizontal direction, for First-order gradient in the vertical direction; The second derivative of the seabed DEM was calculated to obtain the curvature. : .

[0024] The object-level geomorphic attributes of this invention include, but are not limited to, area, perimeter, number of pixels, average elevation, maximum elevation, minimum elevation, elevation standard deviation, local relative elevation difference, average slope, maximum slope, slope standard deviation, average curvature, average absolute curvature, curvature standard deviation, average terrain abrupt change index, maximum terrain abrupt change index, object center elevation, object edge elevation and center-edge elevation difference, aspect ratio, compactness, principal axis direction and other shape attributes.

[0025] This invention calculates the Euclidean distance from each pixel within a non-abrupt region to the nearest topographic abrupt change zone: ; in, Represents any pixel within a non-mutation region. Represents pixels within abrupt topographic changes. This represents the set of pixels representing abrupt changes in terrain. and( ) represent pixels respectively and pixels The row and column coordinates; Represents a cell The Euclidean distance to the nearest topographic abrupt change zone pixel.

[0026] When expressed by raster row and column numbers, the Euclidean distance can be written as: ; in, This indicates the row and column positions of the pixels to be calculated within the non-mutation region. This indicates the row and column positions of pixels in the terrain change zone.

[0027] If we consider the actual pixel resolution of the DEM in both directions, the Euclidean distance can be written as: ; in, and These represent the pixel resolution of the DEM in the horizontal and vertical directions, respectively. The larger the value, the farther the pixel is from the terrain abrupt change zone, and usually closer to the interior of the candidate landform object; The smaller the value, the closer the pixel is to a terrain change zone.

[0028] To reduce local noise and small-scale pseudo-objects caused by over-segmentation, a minimum object size threshold is set after segmenting candidate landform objects. This invention sets the preset noise threshold to 30 pixels, meaning that candidate landform objects with fewer than 30 pixels are considered noise objects and are removed, while areas with at least 30 pixels are retained as valid candidate landform objects.

[0029] For each candidate landform object retained after removing noisy objects, the connected component labeling method is first used to identify mutually independent object regions, and the number of pixels contained in each object is counted, with the number of pixels used as the object area. Then, the objects are sorted in descending order of area, and renumbered using a continuous integer encoding method based on the sorting results. That is, the candidate object with the largest area is numbered 1, the candidate object with the second largest area is numbered 2, and so on, until all candidate landform objects are numbered, finally obtaining the candidate landform object set.

[0030] The GIS vector data formats of this invention include GeoJSON, Shapefile, or other GIS vector data formats, which are used for seabed topographic mapping, spatial statistical analysis, and marine geological interpretation.

[0031] This invention provides a method for identifying landform units by obtaining artificial polygon annotation files corresponding to seabed DEM (Digital Elevation Model) tiles. These files record the categories and boundary point coordinates of landform units such as seamounts, hills, depressions, ridges, and valleys. Since manual annotation is typically based on rendered images or high-resolution images, the image size may differ from the size of the seabed DEM tiles. Therefore, this invention calculates a coordinate scaling ratio based on the size of the annotated image and the DEM tile size, transforming the coordinates of the artificial polygon annotations to the raster coordinate system of the candidate object segmentation results, generating a category annotation mask. Subsequently, the candidate landform objects are spatially superimposed with the category annotation mask, and the number of overlapping pixels between each candidate object and different annotation categories is counted. When the overlap ratio of a certain category in a candidate object exceeds a preset threshold, the candidate object is assigned the corresponding category label. If the same candidate object overlaps with multiple categories, the category with the largest number of overlapping pixels is used as the training label for that candidate object. This step automatically converts artificial polygon annotations into object-level machine learning training samples, avoiding manual reprocessing of object-level samples and improving sample construction efficiency.

[0032] A seafloor topographic unit classification model is trained using the object-level topographic attributes of candidate landform objects as input features and the object categories obtained through manual annotation as output labels. In a preferred embodiment, the classification model is a random forest classification model. A random forest consists of multiple decision trees. Multiple weak classifiers are established by randomly sampling training samples and features, and the final category is output through voting. The random forest model can adapt to multi-dimensional object-level topographic attributes and output the importance of each feature to the classification result. The seafloor topographic unit categories include at least one of seamounts, sea hills, depressions, ridges, valleys, and plains. After training, a classification model for automatic identification of seafloor topographic units is obtained.

[0033] For the seabed DEM data of the area to be identified, topographic factors are calculated, a topographic abrupt change index is constructed, candidate landform objects are segmented, and object-level landform attributes are extracted. The object-level landform attributes of each candidate landform object are input into the trained classification model to obtain the landform unit category and classification confidence score corresponding to the candidate landform object. This yields the object-level seabed topographic unit identification results.

[0034] Based on the raster IDs and classification results of candidate landform objects, a classification raster map of seafloor topographic units is generated. Simultaneously, the boundaries of the candidate landform objects are converted into vector boundaries with geographic coordinate attributes, forming vectorized recognition results. These vectorized recognition results include attributes such as object ID, landform unit category, area, perimeter, local relative elevation difference, average slope, center-edge elevation difference, aspect ratio, compactness, and spatial boundary. The output results can be saved as GeoJSON, Shapefile, or other GIS vector data formats for use in seafloor topographic mapping, spatial statistical analysis, and marine geological interpretation.

[0035] This invention provides a method for identifying seabed topographic units, which can construct a seabed topographic unit identification system. The system includes a DEM data preprocessing module, a topographic factor calculation module, a topographic abrupt change index construction module, a candidate topographic object segmentation module, an object-level topographic attribute extraction module, an object-level sample construction module, a topographic unit classification module, and a spatialization output module. The DEM data preprocessing module reads seabed digital elevation model (DEM) data and performs outlier removal, invalid value imputation, pruning, resampling, and normalization, outputting preprocessed seabed DEM data. The topographic factor calculation module calculates topographic factors such as elevation gradient, slope, slope change rate, and curvature based on the seabed DEM, providing basic data for topographic abrupt change index construction and topographic attribute extraction. The topographic abrupt change index construction module constructs a topographic abrupt change index map based on elevation gradient, slope change rate, and curvature, representing seabed topographic boundaries and morphological transformation areas. The candidate topographic object segmentation module extracts topographic abrupt change zones from the topographic abrupt change index map and segments candidate topographic objects based on topographic abrupt change zone constraints, outputting candidate object number raster. The object-level geomorphic attribute extraction module extracts object-level geomorphic attributes such as elevation, slope, curvature, topographic abrupt change index, shape, and center-edge elevation difference for each candidate geomorphic object. The object-level sample construction module spatially overlays artificial polygon annotations with candidate geomorphic objects to determine the category labels of the candidate objects and generate object-level machine learning training samples. The terrain unit classification module trains a machine learning classification model based on object-level geomorphic attributes and uses the trained model to determine the category of the candidate geomorphic objects to be identified. The spatialization output module converts the classified candidate geomorphic objects into classification raster and vector boundary results, and outputs the category, area, perimeter, geomorphic attributes, and spatial coordinate information.

[0036] The following description, in conjunction with the accompanying drawings and embodiments, provides further details. The method flow of this invention is as follows: Figure 1 As shown, first, DEM data is input and preprocessed, then terrain factors are calculated, a terrain abrupt change index is constructed based on the calculation results, and terrain abrupt change zones are extracted. Candidate landform objects are segmented based on the terrain abrupt change zones, and object-level landform attributes are extracted. Finally, a vectorized segmentation result is formed.

[0037] In this embodiment of the invention, a certain sea area is selected, and six map tiles are extracted, including tile 0001, tile 0039, tile 0030, tile 0061, tile 0083, and tile 0091. Taking tile 0001 as an example, the method of the present invention is used for processing. Figure 2 The input seabed DEM data is used to characterize the elevation undulation and basic geomorphological features of the seabed topography in the study area. It serves as the basic data for subsequent topographic factor calculation, topographic abrupt change index construction, and candidate object segmentation. Figure 3 This is a topographic abrupt change index map constructed from topographic factors such as elevation gradient, slope change rate, and curvature. Brighter areas in the map indicate stronger topographic changes, typically corresponding to seafloor boundaries, slope break lines, or morphological transition zones; darker areas indicate relatively gentler topographic changes. Figure 4 This image shows terrain abrupt change zones extracted based on the terrain abrupt change index. This result is used to characterize potential geomorphic boundaries and as a constraint for segmenting candidate geomorphic objects, reducing problems such as boundary discontinuities, object adhesion, or over-segmentation in ordinary segmentation methods. Figure 5 This is a map showing the candidate geomorphic objects generated based on topographic abrupt change zone constraints. Different colors represent different candidate objects, indicating that continuous DEM raster data has been divided into multiple object-level geomorphic units, providing basic analytical units for subsequent object-level attribute extraction and classification. Figure 6 This is the classification raster result for candidate landform objects. The classification model identifies landform units such as seamounts, sea hills, depressions, ridges, valleys, and plains based on attributes such as elevation, slope, curvature, topographic abruptness index, center-edge elevation difference, aspect ratio, and compactness. Different colors represent different landform unit categories. Figure 7 This image shows the overlay of classification results and object boundaries. It visually illustrates the correspondence between the identified topographic unit boundaries and the original seabed topographic background, and can be used to check the segmentation effect of candidate objects, the rationality of classification results, and the spatial distribution characteristics of different geomorphic units. Seabed topographic units were classified for tiles 0001, 0039, 0030, 0061, 0083, and 0091, and the results are as follows. Figures 8 to 13 As shown, seamounts, hills, depressions, ridges, valleys, and plains are clearly distinguished, with distinct boundaries.

[0038] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for processing seabed topography based on abrupt topographic change zones and object-level geomorphic attributes, characterized in that, include: S1. Determine the monitoring area, acquire the seabed DEM of the monitoring area and preprocess it, and calculate the seabed topographic factor based on the preprocessing results. S2. Construct a topographic abrupt change index map based on seafloor topographic factors, and extract topographic abrupt change zones using the threshold method; use the topographic abrupt change zones as object boundary constraints, and generate candidate landform objects using connected component analysis, distance transformation, and watershed segmentation methods; extract object-level landform attributes based on the spatial extent of the candidate landform objects. S3. Generate an object number raster map based on the raster number of the candidate landform objects, and convert the boundary of each candidate landform object into a vector boundary with geographic coordinate attributes to form a vectorized segmentation result.

2. The seabed topography processing method based on topographic abrupt change zones and object-level geomorphic attributes according to claim 1, characterized in that, S1 includes preprocessing such as outlier removal, invalid value imputation, spatial pruning, resampling, and normalization; the seafloor topographic factors include at least the elevation gradient. ,slope Slope change rate and curvature ; based on Construct a slope map based on Construct a curvature graph.

3. The seabed topography processing method based on topographic abrupt change zones and object-level geomorphic attributes according to claim 2, characterized in that, S2 includes S2.1, the terrain abrupt change index. Depend on , and In summary: ； ； In the formula, For normalization function, , , These are the weighting coefficients.

4. The seabed topography processing method based on topographic abrupt change zones and object-level geomorphic attributes according to claim 3, characterized in that, S2 includes, S2.2, using quantile thresholding or adaptive thresholding methods to... Areas exceeding a threshold are defined as terrain abrupt change zones. Regions less than or equal to the threshold are defined as non-mutation regions, and these non-mutation regions are considered as internal regions of candidate landform objects.

5. The seabed topography processing method based on topographic abrupt change zones and object-level geomorphic attributes according to claim 4, characterized in that, S2 includes S2.3, and the process of generating candidate landform objects includes: Using abrupt terrain change zones as object boundary constraints, connected component analysis is performed on non-abrupt regions to extract all connected pixel components, and these connected pixel components are used as regions to be segmented. For each region to be segmented, perform a distance transformation, calculate the Euclidean distance from each pixel to the nearest terrain abrupt change zone, and obtain the internal distance map of the region; Local maxima points are identified based on distance maps and used as segmentation markers in the watershed segmentation method; The watershed segmentation method is used to divide the current area to be segmented into several candidate landform objects; Calculate the area of ​​each candidate landform object. After segmenting the candidate landform objects, set a minimum object size threshold. Candidate landform objects with fewer than the minimum object size threshold are treated as noise objects and removed. Candidate landform objects with more than or equal to the minimum object size threshold are retained as valid candidate landform objects. For valid candidate landform objects, sort them from largest to smallest area, and renumber them using a continuous integer coding method based on the sorting results.

6. The seabed topography processing method based on topographic abrupt change zones and object-level geomorphic attributes according to claim 5, characterized in that, S2 includes S2.4, which extracts object-level geomorphic attributes for each candidate geomorphic object in the candidate geomorphic object set based on the spatial range of the candidate geomorphic object in the seabed DEM, slope map, curvature map and topographic abrupt change index map.

7. The seabed topography processing method based on topographic abrupt change zones and object-level geomorphic attributes according to claim 6, characterized in that, S3 includes the vectorized segmentation results, which include candidate geomorphic object numbers, geometric boundaries, and object-level geomorphic attributes. The output results are saved in GIS vector data format.

8. The seabed topography processing method based on topographic abrupt change zones and object-level geomorphic attributes according to claim 6, characterized in that, Object-level geomorphic attributes include at least the center-edge elevation difference, aspect ratio, compactness, and topographic abrupt change index.

Images