A verifiable cryptographic fusion method of electronic navigation chart and satellite image

Abstract

This invention discloses a verifiable cryptographic fusion method for electronic nautical charts and satellite imagery, belonging to the field of marine data processing technology. The method includes: preprocessing and registering S-57 electronic nautical chart data and satellite imagery data; constructing a dynamic tidal buffer analysis region; generating a nearshore waters mask with semantic consistency; constructing Merkle tree commitments for the purified electronic nautical chart features and water index raster; extracting nearshore boundaries and calculating geometric consistency scores; calculating semantic consistency scores based on the nearshore waters mask and depth region reference maps; and finally, weighted fusion of geometric, semantic, and commitment scores to generate a reliable fusion index, which is then classified and evaluated according to a threshold. This invention combines geometric consistency, semantic consistency, and cryptographic integrity verification, enabling quantifiable and verifiable evaluation of the fusion results of multi-source marine geospatial data, providing a standardized technical path for hydrological quality assurance and chart updates.

Description

Technical Field

[0001] This invention relates to the field of marine data processing technology, and in particular to a verifiable cryptographic fusion method for electronic nautical charts and satellite imagery. Background Technology

[0002] Electronic nautical charts are the official digital versions of paper nautical charts, developed according to the International Hydrographic Organization's S-57 standard. They encode key navigational safety elements such as coastlines, water depth data, and navigational aids, serving as the fundamental data source for electronic chart display and information systems. Meanwhile, high-resolution multispectral satellite imagery, exemplified by Sentinel-2, has been widely applied in areas such as coastline extraction, shallow water depth retrieval, and nearshore environmental monitoring due to its advantages of wide coverage, short revisit cycles, and low acquisition costs. The fusion analysis of electronic nautical charts and satellite imagery has significant application value for assessing chart timeliness, detecting coastline changes, and updating hydrological elements. However, existing technologies have the following limitations in achieving the fusion assessment of electronic nautical charts and satellite imagery: The current evaluation dimensions are limited, with research focusing either on geometric consistency, such as distance metrics like coastline displacement and boundary morphology differences, or on semantic performance, such as confusion matrix metrics like intersection-over-union ratio and F1 score for pixel-level classification. Joint evaluation of geometric and semantic consistency within a unified framework is rare and fails to comprehensively reflect the overall quality of fused products.

[0003] The lack of integrity verification has exacerbated the risk of data tampering as geospatial data is generated, transmitted, and stored in distributed network systems. Traditional precision-oriented assessment metrics cannot answer key questions such as whether fused products have been tampered with and their reliability. Existing data protection schemes such as S-63 mainly regulate encryption and authentication during the distribution process, failing to provide fine-grained integrity assessments for derived fused products.

[0004] The lack of standardized decision-making basis means that the existing marine geographic information system fusion framework relies heavily on heterogeneous quality indicators, making it difficult to compare across products and regions, which limits its practical application in automated chart maintenance and risk perception hydrological quality assurance processes.

[0005] To address the aforementioned issues, there is an urgent need for a method for fusing electronic nautical charts and satellite imagery that can jointly assess geometric and semantic consistency and possess cryptographic verifiability, thus filling the gaps in existing technologies. Therefore, this paper proposes a verifiable cryptographic fusion method for electronic nautical charts and satellite imagery to solve the aforementioned problems. Summary of the Invention

[0006] The purpose of this invention is to provide a verifiable cryptographic fusion method for electronic nautical charts and satellite imagery to solve the problems mentioned in the background art.

[0007] This invention is achieved through the following technical solution: A verifiable cryptographic fusion method for electronic nautical charts and satellite imagery, the method comprising the following steps: Step 1: Acquire electronic nautical chart S-57 data and satellite imagery data of the target area, preprocess and register the electronic nautical chart S-57 data and satellite imagery data, and unify them to a preset spatial reference system; Step 2: Based on the preprocessed electronic nautical chart data, combined with tidal information and uncertainties, construct a dynamic tidal buffer analysis area; Step 3: Within the tidal buffer analysis area, calculate the water index based on the preprocessed satellite image data, and combine it with the depth area information in the electronic nautical chart to generate a nearshore water mask with semantic consistency. Step 4: Clean up the vector geometry in the preprocessed electronic nautical chart data, and divide the cleaned electronic nautical chart features and the water index raster calculated in Step 3 into blocks, construct cryptographic Merkle tree commitments respectively, and generate commitment components. Step 5: Based on the nearshore waters mask and electronic nautical chart data, extract their respective nearshore boundaries, and obtain the geometric consistency score by calculating the distance metric of the boundary point set; Step 6: Based on the nearshore waters mask and the rasterized electronic nautical chart depth region reference map, calculate the pixel-level semantic segmentation index to obtain the semantic consistency score; Step 7: The geometric consistency score, semantic consistency score and commitment component are weighted and fused to generate a scalar trust fusion index, and the fusion result is classified and evaluated according to a preset decision threshold.

[0008] Furthermore, the preprocessing and registration of the electronic nautical chart S-57 data in step 1 further includes: converting the S-57 data into a GeoPackage format geographic package and reprojecting it to a preset universal transverse Mercator projection coordinate system; and the preprocessing and registration of the satellite image data further includes: importing, stitching, resampling, and reprojecting the satellite image to the same universal transverse Mercator projection coordinate system as the electronic nautical chart data, generating raster data in cloud-optimized GeoTIFF format.

[0009] Furthermore, in step 2, constructing a dynamic tidal buffer analysis region specifically involves: using the coverage area of ​​the electronic nautical chart as the base region, calculating the horizontal displacement caused by tides and its related uncertainties, and using this to expand the preset base buffer radius to obtain an effective buffer radius. The base region is then expanded using the effective buffer radius to obtain the dynamic tidal buffer analysis region.

[0010] Furthermore, in step 3, generating a nearshore water mask with semantic consistency further includes: Within the tidal buffer analysis area, calculate the normalized difference water index and / or the improved normalized difference water index of satellite imagery; The optimal water body index segmentation threshold is automatically determined through a constraint optimization process. The constraint optimization process takes the water body pixels after the depth region polygon rasterization in the electronic nautical chart as a reference. By setting a minimum recall constraint, the pixels of the water body marked on the nautical chart are correctly identified, while optimizing the balance accuracy of water body and non-water body classification. The optimal threshold is applied to the water index to generate an initial binary water area mask, which is then spatially post-processed to maintain connectivity with the depth region of the electronic nautical chart, ultimately yielding the nearshore water area mask.

[0011] Furthermore, in step 4, constructing the cryptographic Merkle tree commitment further includes: The vector geometry in the electronic nautical chart is cleaned up, including repairing invalid geometry, unifying coordinate precision, and serializing it into a canonical byte sequence. Leaf node hash values ​​are generated through a hash function, and Merkle trees are constructed by grouping by layer to generate root hash as a commitment of the electronic nautical chart dataset. The water index raster calculated in step 3 is divided into fixed-size blocks. The pixel value of each tile is standardized and serialized into a byte sequence. The leaf node hash value is generated by a hash function, and a Merkle tree is constructed to generate the root hash as the commitment of the water index raster. The commitment component is calculated by weighting the leaf-level pass rate of the electronic nautical chart dataset and the tile-level pass rate of the water index raster.

[0012] Furthermore, in step 5, calculating the geometric consistency score further includes: Extract image boundary point sets from nearshore water masks, and extract chart boundary point sets from coastline features and / or depth area boundaries on electronic nautical charts; Calculate the bidirectional chamfer distance and bidirectional Hausdorff distance between two sets of boundary points, and normalize the distance metric to the interval of 0 to 1 using an exponential decay function; Based on the preset tolerance radius, calculate the boundary F1 score between two boundary point sets; The normalized distance metric is weighted and combined with the boundary F1 score to obtain the final geometric consistency score.

[0013] Furthermore, in step 6, a semantic consistency score is obtained, specifically by calculating the pixel-level classification F1 score between the nearshore waters mask determined in step 3 and the rasterized electronic nautical chart depth area reference map, and using the F1 score as the semantic consistency score.

[0014] Furthermore, in step 7, the geometric consistency score, semantic consistency score, and commitment component are weighted and fused. Specifically, a preset weight is assigned to each of the geometric consistency, semantic consistency, and commitment component, and the sum of the three weights is 1. A scalar in the range of 0 to 1 is obtained by weighted summation, which is the trustworthy fusion index.

[0015] Furthermore, in step 7, the fusion result is classified and evaluated according to a preset decision threshold, specifically: a higher first threshold and a lower second threshold are set; if the trustworthy fusion index is greater than or equal to the first threshold, the fusion result is determined to be trustworthy; if the trustworthy fusion index is between the first threshold and the second threshold, the fusion result is determined to be under review; if the trustworthy fusion index is less than the second threshold, the fusion result is determined to be abnormal.

[0016] Furthermore, the satellite imagery data is Sentinel-2 Level-2A data; the preset spatial reference system is the universal transverse Mercator projection coordinate system; the preset geographic package is in GeoPackage format; and the cloud-optimized geotagged image file format is in cloud-optimized GeoTIFF format.

[0017] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: 1. This invention integrates traditional geometric accuracy indicators, pixel-level classification indicators, and cryptographic integrity verification indicators into the same framework by constructing geometric consistency scoring, semantic consistency scoring, and commitment consistency components. This overcomes the shortcomings of existing technologies that have a single evaluation dimension and can comprehensively and objectively reflect the overall quality of electronic nautical chart and satellite image fusion products.

[0018] 2. This invention employs Merkle tree commitment technology to construct cryptographic commitments for both electronic nautical chart vector features and satellite imagery water index grids, enabling the fused product to possess post-event verifiable tamper detection capabilities. The commitment component C can sensitively respond to data modifications and identify integrity violations that cannot be detected by purely geometric-semantic indicators, providing technical assurance for data traceability and trustworthiness assessment.

[0019] 3. This invention integrates multi-dimensional evaluation results into a scalar-form credibility fusion index and sets an interpretable decision threshold, classifying the fusion results into three states: credible, under review, and abnormal. This standardized output can directly serve operational scenarios such as hydrological quality assurance, priority ranking of nautical chart updates, and automatic screening of suspicious areas, demonstrating good practicality and scalability.

[0020] 4. The tidal buffer analysis region construction method of the present invention takes into account tidal displacement and measurement uncertainty, making the evaluation insensitive to reasonable changes in the analysis range; the automatic threshold selection mechanism based on constraint optimization can adapt to the spectral characteristics of different water bodies; cross-regional experiments show that the present invention can maintain stable evaluation performance under different geomorphic features and different water body conditions. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only preferred embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating a verifiable cryptographic fusion method for electronic nautical charts and satellite imagery provided by the present invention.

[0023] Figure 2 This is a flowchart illustrating the four-layer trusted fusion framework of the present invention.

[0024] Figure 3 This is a visual alignment check result of Sentinel-2 true-color image and ENC coastline in Yazhou Bay in an embodiment of the present invention.

[0025] Figure 4 This is a comparison chart of confusion matrices and semantic segmentation performance indicators for three hydrological regions—Yazhou Bay, Dongwan River section, and Potomac River section—in this embodiment of the invention.

[0026] Figure 5 The figures show the measurement results of the reliable fusion index of three hydrological regions and the comparison of G, S, and C components in this embodiment of the invention.

[0027] Figure 6 This is a schematic diagram of the spatial overlay of ENC-Sentinel-2 for three hydrological regions in an embodiment of the present invention.

[0028] Figure 7 This is a sensitivity analysis diagram of the NDWI / MNDWI threshold on the F1 score and the confidence fusion index in an embodiment of the present invention.

[0029] Figure 8 This is a time evolution diagram of G, S, C and the credible fusion index from four Sentinel-2 multi-temporal observations in Yazhou Bay in an embodiment of the present invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the present invention more apparent, exemplary embodiments according to the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely a part of the embodiments of the present invention, and not all of the embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein. Based on the embodiments of the present invention described herein, all other embodiments obtained by those skilled in the art without inventive effort should fall within the protection scope of the present invention.

[0031] See Figure 1-8 A verifiable cryptographic fusion method for electronic nautical charts and satellite imagery includes the following steps: Step 1: Acquire electronic nautical chart S-57 data and satellite imagery data of the target area, preprocess and register the electronic nautical chart S-57 data and satellite imagery data, and unify them to a preset spatial reference system; Step 2: Based on the preprocessed electronic nautical chart data, combined with tidal information and uncertainties, construct a dynamic tidal buffer analysis area; Step 3: Within the tidal buffer analysis area, calculate the water index based on the preprocessed satellite image data, and combine it with the depth area information in the electronic nautical chart to generate a nearshore water mask with semantic consistency. Step 4: Clean up the vector geometry in the preprocessed electronic nautical chart data, and divide the cleaned electronic nautical chart features and the water index raster calculated in Step 3 into blocks, construct cryptographic Merkle tree commitments respectively, and generate commitment components. Step 5: Based on the nearshore waters mask and electronic nautical chart data, extract their respective nearshore boundaries, and obtain the geometric consistency score by calculating the distance metric of the boundary point set; Step 6: Based on the nearshore waters mask and the rasterized electronic nautical chart depth region reference map, calculate the pixel-level semantic segmentation index to obtain the semantic consistency score; Step 7: The geometric consistency score, semantic consistency score and commitment component are weighted and fused to generate a scalar trust fusion index, and the fusion result is classified and evaluated according to a preset decision threshold.

[0032] Step 1, which involves preprocessing and registering the electronic nautical chart S-57 data, further includes: converting the S-57 data into a GeoPackage format geographic package and reprojecting it to a preset universal transverse Mercator projection coordinate system; and preprocessing and registering the satellite image data, which further includes: importing, stitching, resampling, and reprojecting the satellite image to the same universal transverse Mercator projection coordinate system as the electronic nautical chart data, generating raster data in cloud-optimized GeoTIFF format.

[0033] In step 2, constructing a dynamic tidal buffer analysis region specifically involves: using the coverage area of ​​the electronic nautical chart as the base region, calculating the horizontal displacement caused by tides and its related uncertainties, and using this to expand the preset base buffer radius to obtain an effective buffer radius. The base region is then expanded using the effective buffer radius to obtain the dynamic tidal buffer analysis region.

[0034] Step 3, generating a semantically consistent nearshore water mask, further includes: Within the tidal buffer analysis area, calculate the normalized difference water index and / or the improved normalized difference water index of satellite imagery; The optimal water body index segmentation threshold is automatically determined through a constraint optimization process. The constraint optimization process takes the water body pixels after the depth region polygon rasterization in the electronic nautical chart as a reference. By setting a minimum recall constraint, the pixels of the water body marked on the nautical chart are correctly identified, while optimizing the balance accuracy of water body and non-water body classification. The optimal threshold is applied to the water index to generate an initial binary water area mask, which is then spatially post-processed to maintain connectivity with the depth region of the electronic nautical chart, ultimately yielding the nearshore water area mask.

[0035] Step 4, constructing a cryptographic Merkle tree commitment, further includes: The vector geometry in the electronic nautical chart is cleaned up, including repairing invalid geometry, unifying coordinate precision, and serializing it into a canonical byte sequence. Leaf node hash values ​​are generated through a hash function, and Merkle trees are constructed by grouping by layer to generate root hash as a commitment of the electronic nautical chart dataset. The water index raster calculated in step 3 is divided into fixed-size blocks. The pixel value of each tile is standardized and serialized into a byte sequence. The leaf node hash value is generated by a hash function, and a Merkle tree is constructed to generate the root hash as the commitment of the water index raster. The commitment component is calculated by weighting the leaf-level pass rate of the electronic nautical chart dataset and the tile-level pass rate of the water index raster.

[0036] Step 5, calculating the geometric consistency score, further includes: Extract image boundary point sets from nearshore water masks, and extract chart boundary point sets from coastline features and / or depth area boundaries on electronic nautical charts; Calculate the bidirectional chamfer distance and bidirectional Hausdorff distance between two sets of boundary points, and normalize the distance metric to the interval of 0 to 1 using an exponential decay function; Based on the preset tolerance radius, calculate the boundary F1 score between two boundary point sets; The normalized distance metric is weighted and combined with the boundary F1 score to obtain the final geometric consistency score.

[0037] In step 6, a semantic consistency score is obtained, specifically by calculating the F1 score for pixel-level classification between the nearshore waters mask determined in step 3 and the rasterized electronic nautical chart depth area reference map, and using the F1 score as the semantic consistency score.

[0038] In step 7, the geometric consistency score, semantic consistency score, and commitment component are weighted and fused. Specifically, a preset weight is assigned to each of the geometric consistency, semantic consistency, and commitment component, and the sum of the three weights is 1. A scalar in the range of 0 to 1 is obtained by weighted summation, which is the trust fusion index.

[0039] In step 7, the fusion result is classified and evaluated according to a preset decision threshold. Specifically, a higher first threshold and a lower second threshold are set. If the credibility fusion index is greater than or equal to the first threshold, the fusion result is determined to be in a credible state. If the credibility fusion index is between the first threshold and the second threshold, the fusion result is determined to be in a review state. If the credibility fusion index is less than the second threshold, the fusion result is determined to be in an abnormal state.

[0040] The satellite imagery data is Sentinel-2 Level-2A data; the preset spatial reference system is the universal transverse Mercator projection coordinate system; the preset geographic package is GeoPackage format; and the cloud-optimized geotagged image file format is cloud-optimized GeoTIFF format.

[0041] For example, based on the application of the method of the present invention in the Yazhou Bay area, this embodiment takes Yazhou Bay in Sanya City, Hainan Province, China as the research area, and uses the official S-57 electronic nautical chart unit covering the area and Sentinel-2 Level-2A satellite imagery as the data source to fully implement and verify the method of the present invention.

[0042] Yazhou Bay is located on the southwestern coast of Sanya City, Hainan Province, China, where the Ningyuan River flows into the bay. The coastline of this area includes beaches, ports, and emerging urban and industrial zones associated with the Yazhou Bay Science and Technology City, creating a mixed natural and man-made environment. The nearshore area is relatively shallow and gentle, featuring tidal flats, intertidal beaches, and dredged channels. The tidal type is the typical regular diurnal tide of the northern South China Sea, with moderate tidal range. This heterogeneous morphodynamic characteristic of the coexistence of natural coastlines, port infrastructure, and rapidly developing coastal facilities constitutes an ideal testing ground for evaluating the consistency between ENC and satellite imagery. The electronic nautical chart data uses the official S-57 ENC unit covering Yazhou Bay and its surrounding waters, including standard hydrological and mapping object categories such as coastline COALNE, land area LNDARE, depth area DEPARE, depth contour lines DEPCNT, water depth data SOUNDG, navigation aids, and various annotation features. This ENC, as an authoritative nautical chart, serves a dual function of geometric and semantic reference. Satellite imagery data was obtained using the Sentinel-2 Level-2A surface albedo product. Cloudless or low-cloud scenes were selected. The green band B03 (10 meters), red band B04 (10 meters), blue band B02 (10 meters), and near-infrared band B08 (10 meters) were used for true-color visualization and NDWI calculation. The 20-meter shortwave infrared band B11 was resampled and used for MNDWI calculation.

[0043] Step 1: Preprocessing and Registration The Yazhou Bay S-57 ENC unit was imported into the system, converted into a GeoPackage format geographic package, and reprojected to the UTM 49N coordinate system (EPSG:32649). All feature classes were stored as multi-part geometries with spatial indexes. Simultaneously, the Sentinel-2 Level-2A SAFE product of the corresponding time phase was imported, and the blue, green, red, and near-infrared bands at 10-meter resolution were stitched together and reprojected to the same UTM 49N area, generating cloud-optimized GeoTIFF raster data. Since Level-2A surface reflectance is stored as scaled integers, it was converted to dimensionless atmospheric sub-atmosphere reflectance by dividing by 10000 and cropping the values ​​to a reasonable range. A robust linear stretching mapping was used to map the central part of the reflectance distribution to the [0,1] interval to reduce the impact of extreme anomalies such as residual clouds or solar reflection.

[0044] The alignment verification module quantifies CRS consistency, pixel size, and rotation angle to calculate the spatial overlap between the cropped Sentinel-2 image coverage and the ENC coastline coverage. A three-level status (pass / warning / fail) is generated, allowing only scenes with a satisfactory alignment status to proceed to subsequent processing.

[0045] Step 2: Construct the dynamic tidal buffer analysis region Based on the coverage area of ​​the electronic nautical chart, the initial analysis domain is derived from the ENC coverage object M_COVR, and supplemented by the convex hull of the merged ENC geometry (the union of LNDARE and DEPARE / SEAARE). Tidal prediction data for the Yazhou Bay area are acquired, and the tidal height difference between the imaging time and the chart reference level is calculated. Combined with near-shore slope Estimate the horizontal displacement caused by tides The total uncertainty is obtained by aggregating the contributions of uncertainties such as ENC location tolerance, image registration error, tidal prediction uncertainty, and seasonal morphological changes using the root mean square method. Set the base buffer radius. 2000 meters, region-specific expansion factor Taking 1.0 (reflecting the moderate coastal complexity of Yazhou Bay), the formula is used... The calculated effective buffer radius is approximately 2056.7 meters. This effective buffer radius is used to expand the base area, resulting in a dynamic tidal buffer analysis region. Sentinel-2 imagery is then cropped into this analysis region to generate a simplified scene focused on the nearshore area.

[0046] Step 3: Generate nearshore water mask Calculate the normalized difference water index on the cropped Sentinel-2 image:

[0047] Corrected Normalized Difference Water Index:

[0048] The DEPARE depth region in the electronic nautical chart is rasterized into polygons and used as a reference water area mask. The NDWI threshold is traversed within the range [-0.2, 0.4] with a step size of 0.025, and the corresponding MNDWI grid is adjusted synchronously. For each candidate threshold... Calculate the confusion matrix with the reference mask to obtain the recall rate. And balance accuracy Set a minimum recall rate constraint. In order to satisfy Among the thresholds, the combination that maximizes the balance accuracy is selected, and the NDWI threshold is determined to be 0.10 and the MNDWI threshold to be 0.08. This threshold is applied to generate an initial binary water area mask, and spatial post-processing is performed: isolated connected regions with an area less than 10 pixels are removed to suppress speckled artifacts, and connectivity with the nautical-annotated water areas is ensured by intersecting with the dilated DEPARE mask, ultimately yielding the nearshore water area mask.

[0049] Step 4: Constructing a cryptographic Merkle tree commitment The vector geometry of the electronic nautical chart undergoes purification: invalid polygons are repaired, polygon orientations are unified, and Z-coordinates are removed to obtain a pure 2D geometry. The coordinates are converted to a fixed decimal representation with 0.01-meter precision (far exceeding the 10-meter pixel size of Sentinel-2 and typical ENC positioning accuracy). The purified geometry is encoded in a standard Well-Known Binary format, and relevant attributes are serialized into a standard JSON string with deterministic key order. A unique byte sequence is generated for each feature, and leaf node hash values ​​are generated using SHA-256 hashing. The leaf nodes are grouped according to the object category code OBJL, and a Merkle tree is constructed for each group. The ordered connections of the hash values ​​of the recursively hashed child nodes are used to generate a root hash as a commitment for the ENC dataset.

[0050] The NDWI grid obtained in step 3 is divided into fixed-size blocks of 64×64 pixels. The floating-point values ​​of each tile are standardized into a uniform memory representation, and missing values ​​are replaced with predefined values. Sentinel values, generated as a continuous byte array, are hashed using SHA-256 to form the leaf nodes of a Merkle tree. The root hash is recursively generated as a commitment for the NDWI raster.

[0051] The storage commitment components include the Merkle root, leaf node hashes, and metadata. During subsequent verification, the hashes of the current GeoPackage and NDWI raster are recalculated and compared with the storage commitments to calculate the leaf-level pass rate. and tile-level pass rate Receive commitment components In this embodiment, all commitments passed verification, and C=1.000.

[0052] Step 5: Calculate the geometric consistency score The land-water interface is refined from the nearshore water mask and converted into a polyline. Equal-interval sampling is then performed to extract the image boundary point set. From the electronic nautical chart, the boundary edges of the COALNE line segment and the LNDARE and DEPARE / SEAARE polygons are merged to construct the chart boundary point set, explicitly excluding internal boundaries between depth regions.

[0053] Build Tree-based calculation of symmetric nearest neighbor distance between two point sets; derivation of bidirectional chamfered distance. Meters, 95th percentile Hausdorf distance in both directions Meters. Distance measurement is passed through an exponential decay function. Normalization Take 20 meters (approximately 2 pixels). get

[0054] .

[0055] The boundary F1 score was calculated based on a set of tolerance radii [5m, 10m, 15m, 20m, 25m, 30m], and the maximum observed F1 value was determined. .

[0056] set up , , calculate:

[0057]

[0058] Geometric consistency score obtained after calibration (The final value has been adjusted by the system calibration factor to make the score more consistent with visual perception.)

[0059] Step 6: Calculate the semantic consistency score Pixel-level classification confusion matrices and various metrics were calculated between the nearshore water mask and the rasterized DEPARE reference image. Results showed that the intersection-union ratio (IU / I) was... , , , , , , .Will As a semantic consistency score .

[0060] Step 7: Generate a trustworthy fusion index Set weights , , Geometric consistency score Semantic consistency score , Commitment Component

[0061] The credible fusion index is obtained by performing a weighted summation. , Take three significant digits, which is 0.900.

[0062] Set the first threshold Second threshold ,because The fusion result is determined to be reliable. Meanwhile, because... It meets the "trusted / clean" integrity status.

[0063] Ablation experiment verification To verify the contribution of each module, the following ablation experiments were conducted: Ablation Experiment I: Tidal AOI Buffering Effect Comparison of no tidal buffer (only 2000 meters of fixed buffer) and tidal buffer enabled ( Two configurations were used (meters). After enabling the tidal buffer, the NDWI assessment area increased from 598.4 km² to 602.8 km², an increase of approximately 0.73%. The geometric component G remained at 0.763, the semantic component S remained at 0.986, and the commitment component C was 1.000 for all. The TFI differed only in the fourth decimal place (0.89965 vs 0.89964), indicating that the tidal buffer mainly provides a physical safety margin and its impact on global indicators is negligible.

[0064] Ablation Experiment II: Outlier Removal A comparison was made between two configurations: one with outlier removal disabled and the other with it enabled. With removal disabled, the HD95 distance increased from 22.29 meters to 66.64 meters, the average distance increased from 10.38 meters to 26.63 meters, the geometric score G decreased from 0.763 to 0.687, and the TFI decreased from 0.900 to 0.869. The experiments demonstrate that the outlier removal mechanism effectively prevents a very small number of highly inconsistent pixels from dominating global statistics, generating a more stable and interpretable geometric score.

[0065] Ablation Experiment III: Trust Layer Effect Comparison of four settings: Cleaning baseline with trust enabled, Trust layer disabled, Light tampering (2% NDWI tiles), and Heavy tampering (20% NDWI tiles). Cleaning baseline , Minor alteration , This triggers the "Affected (Under Review)" status; severe tampering. , This triggers a "damaged / abnormal" state; when the trust layer is disabled, only geometric-semantic scoring is performed. It is not sensitive to tampering. Experiments have shown that the commitment component C can effectively identify integrity violations, while pure geometric-semantic metrics cannot distinguish them.

[0066] Sensitivity analysis verification Sensitivity Analysis I: AOI Radius and Tidal Factor Tests were conducted under three representative configurations: ( Factor 0.5 ), (Baseline) ( Factor 1.5 The AOI area increased from 558.2 km² to 660.2 km², but G remained in the range of 0.7629-0.7633, S remained stable at 0.9862, and TFI fluctuated between 0.8996 and 0.8998, with a fluctuation range of about 0.02 percentage points, indicating that the method is highly robust to the selection of AOI parameters.

[0067] Sensitivity Analysis II: NDWI / MNDWI Threshold The NDWI threshold was adjusted in steps of 0.025 within the range of [0.025, 0.20], and the MNDWI threshold was adjusted accordingly. The F1 curve formed a broad plateau around the automatically selected thresholds (0.10 / 0.08), and only gradual degradation occurred when deviating from the optimal value of 0.025-0.05. The difference between F1 and TFI was approximately... The study demonstrates that a threshold selection mechanism based on recall constraints and driven by balanced precision can achieve near-optimal semantic performance and is insensitive to small perturbations in the threshold.

[0068] Multi-phase testing Four Sentinel-2 images of Yazhou Bay, representing different seasons, were used for testing, covering the winter low tide period, late spring / early summer, summer rainy season, and the post-autumn storm period. The entire process was repeated for each date, maintaining the same ENC reference and parameter settings. The results show that semantic component S maintains extremely high stability across all scenarios. The geometric component G exhibits moderate variability (range 0.75-0.77), reflecting the true changes in coastline appearance and wave and sedimentary characteristics; the commitment component C is consistently 1.000; and the TFI remains stable between 0.89 and 0.92, all within the confidence range. Experiments demonstrate that the method is time-robust to seasonal and tidal variations.

[0069] Cross-region generalization verification To verify the generalization ability of this invention, two S-57 electronic nautical chart units, US5MD13M (East Bay section) and US5MD41M (Potomac section), and their corresponding Sentinel-2 images from the Chesapeake Bay region of the United States were selected for testing. All parameter settings were kept constant, only the data source was changed. Table 1 shows the NDWI / MNDWI thresholds and water mask performance for the three hydrological regions, and Table 2 shows the geometric / semantic consistency index and TFI score for the three hydrological regions.

[0070] Table 1

[0071] Table 2

[0072] US5MD13M East Bay River Section This area covers the tidal, deep-water zone of the Chesapeake Bay main channel, containing complex bays and channels. The automatically determined NDWI / MNDWI thresholds are 0.04 / 0.06, generating a water mask with F1=98.0% and IoU=96.5%. The geometric consistency score G=0.695 (reflecting a relatively complex shoreline morphology), semantic consistency score S=0.988, commitment component C=1.000, and TFI=0.873, indicating a trustworthy state.

[0073] US5MD41M Potomac River Section This area encompasses the lower reaches of the Potomac River, a meandering channel dotted with islands and narrow tributaries. Due to the river's dark color and high turbidity, the automatic threshold was adjusted to -0.06 / -0.12, F1=99.1%, IoU=98.3%. The geometric consistency score was G=0.841 (reflecting the high degree of fit between the coastline and the water boundary), S=0.991, C=1.000, and TFI=0.933.

[0074] Test results show that the method of this invention can adaptively determine the threshold under different terrain and water conditions, maintain high semantic consistency, and effectively reflect the geometric differences between regions through TFI. The TFI values ​​for the three regions are 0.900, 0.873, and 0.933, respectively, all within the confidence interval, verifying the method's cross-regional generalization ability.

[0075] This invention is implemented in Python and utilizes open-source geospatial libraries such as rasterio, GDAL, GeoPandas, Shapely, and SciPy. The system architecture strictly corresponds to a four-layer framework, employing a modular design, with each step interacting through clearly defined file inputs and outputs.

[0076] Data is organized according to the following standardized directory structure: raw S-57 and Sentinel-2 SAFE data are stored in data / raw / ; standardized geographic packages, COG files, NDWI / MNDWI rasters, water masks, and log files are stored in data / processed / ; Merkle tree commitments and metadata are stored in data / commit / ; fusion evaluation documents, validation summaries, and TFI scores are included in data / verify / ; and charts, overlays, and table summaries are stored in data / results / .

[0077] A lightweight web portal built on Flask and Leaflet supports interactive viewing of contrast-enhanced true-color PNG images, ENC layers (land area, depth area, depth contour lines, coastline, and water depth data), and UTM grids, and displays the current TFI component and classification status in real time via a JSON endpoint.

[0078] This invention provides a method for fusing electronic nautical charts and satellite imagery that can jointly assess geometric and semantic consistency and has cryptographic verifiability. Through a four-layer trusted fusion framework (preprocessing and registration layer, water body segmentation and semantic fusion layer, trust establishment layer, and fusion evaluation and TFI layer) and a scalar trusted fusion index (TFI), this invention achieves quantifiable and verifiable evaluation of the fusion results of multi-source marine geospatial data, providing a standardized and operable technical path for hydrological quality assurance, chart update decision-making, and automated screening of suspicious areas.

[0079] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A verifiable cryptographic fusion method for electronic nautical charts and satellite imagery, characterized in that, The method includes the following steps: Step 1: Acquire electronic nautical chart S-57 data and satellite imagery data of the target area, preprocess and register the electronic nautical chart S-57 data and satellite imagery data, and unify them to a preset spatial reference system; Step 2: Based on the preprocessed electronic nautical chart data, combined with tidal information and uncertainties, construct a dynamic tidal buffer analysis area; Step 3: Within the tidal buffer analysis area, calculate the water index based on the preprocessed satellite image data, and combine it with the depth area information in the electronic nautical chart to generate a nearshore water mask with semantic consistency. Step 4: Clean up the vector geometry in the preprocessed electronic nautical chart data, and divide the cleaned electronic nautical chart features and the water index raster calculated in Step 3 into blocks, construct cryptographic Merkle tree commitments respectively, and generate commitment components. Step 5: Based on the nearshore waters mask and electronic nautical chart data, extract their respective nearshore boundaries, and obtain the geometric consistency score by calculating the distance metric of the boundary point set; Step 6: Based on the nearshore waters mask and the rasterized electronic nautical chart depth region reference map, calculate the pixel-level semantic segmentation index to obtain the semantic consistency score; Step 7: The geometric consistency score, semantic consistency score and commitment component are weighted and fused to generate a scalar trust fusion index, and the fusion result is classified and evaluated according to a preset decision threshold.

2. The verifiable cryptographic fusion method according to claim 1, characterized in that, Step 1, which involves preprocessing and registering the electronic nautical chart S-57 data, further includes: converting the S-57 data into a GeoPackage format geographic package and reprojecting it to a preset universal transverse Mercator projection coordinate system; and preprocessing and registering the satellite image data, which further includes: importing, stitching, resampling, and reprojecting the satellite image to the same universal transverse Mercator projection coordinate system as the electronic nautical chart data, generating raster data in cloud-optimized GeoTIFF format.

3. The verifiable cryptographic fusion method according to claim 1, characterized in that, In step 2, constructing a dynamic tidal buffer analysis region specifically involves: using the coverage area of ​​the electronic nautical chart as the base region, calculating the horizontal displacement caused by tides and its related uncertainties, and using this to expand the preset base buffer radius to obtain an effective buffer radius. The base region is then expanded using the effective buffer radius to obtain the dynamic tidal buffer analysis region.

4. The verifiable cryptographic fusion method according to claim 1, characterized in that, Step 3, generating a semantically consistent nearshore water mask, further includes: Within the tidal buffer analysis area, calculate the normalized difference water index and / or the improved normalized difference water index of satellite imagery; The optimal water body index segmentation threshold is automatically determined through a constraint optimization process. The constraint optimization process takes the water body pixels after the depth region polygon rasterization in the electronic nautical chart as a reference. By setting a minimum recall constraint, the pixels of the water body marked on the nautical chart are correctly identified, while optimizing the balance accuracy of water body and non-water body classification. The optimal threshold is applied to the water index to generate an initial binary water area mask, which is then spatially post-processed to maintain connectivity with the depth region of the electronic nautical chart, ultimately yielding the nearshore water area mask.

5. The verifiable cryptographic fusion method according to claim 1, characterized in that, Step 4, constructing a cryptographic Merkle tree commitment, further includes: The vector geometry in the electronic nautical chart is cleaned up, including repairing invalid geometry, unifying coordinate precision, and serializing it into a canonical byte sequence. Leaf node hash values ​​are generated through a hash function, and Merkle trees are constructed by grouping by layer to generate root hash as a commitment of the electronic nautical chart dataset. The water index raster calculated in step 3 is divided into fixed-size blocks. The pixel value of each tile is standardized and serialized into a byte sequence. The leaf node hash value is generated by a hash function, and a Merkle tree is constructed to generate the root hash as the commitment of the water index raster. The commitment component is calculated by weighting the leaf-level pass rate of the electronic nautical chart dataset and the tile-level pass rate of the water index raster.

6. The verifiable cryptographic fusion method according to claim 1, characterized in that, Step 5, calculating the geometric consistency score, further includes: Extract image boundary point sets from nearshore water masks, and extract chart boundary point sets from coastline features and / or depth area boundaries on electronic nautical charts; Calculate the bidirectional chamfer distance and bidirectional Hausdorff distance between two sets of boundary points, and normalize the distance metric to the interval of 0 to 1 using an exponential decay function; Based on the preset tolerance radius, calculate the boundary F1 score between two boundary point sets; The normalized distance metric is weighted and combined with the boundary F1 score to obtain the final geometric consistency score.

7. The verifiable cryptographic fusion method according to claim 1, characterized in that, In step 6, a semantic consistency score is obtained, specifically by calculating the F1 score for pixel-level classification between the nearshore waters mask determined in step 3 and the rasterized electronic nautical chart depth area reference map, and using the F1 score as the semantic consistency score.

8. The verifiable cryptographic fusion method according to claim 1, characterized in that, In step 7, the geometric consistency score, semantic consistency score, and commitment component are weighted and fused. Specifically, a preset weight is assigned to each of the geometric consistency, semantic consistency, and commitment component, and the sum of the three weights is 1. A scalar in the range of 0 to 1 is obtained by weighted summation, which is the trust fusion index.

9. The verifiable cryptographic fusion method according to claim 1, characterized in that, In step 7, the fusion result is classified and evaluated according to a preset decision threshold. Specifically, a higher first threshold and a lower second threshold are set. If the credible fusion index is greater than or equal to the first threshold, the fusion result is determined to be credible. If the trustworthy fusion index is between the first threshold and the second threshold, the fusion result is determined to be under review. If the credible fusion index is less than the second threshold, the fusion result is determined to be abnormal.

10. The verifiable cryptographic fusion method according to any one of claims 1 to 9, characterized in that, The satellite imagery data is Sentinel-2 Level-2A data; the preset spatial reference system is the universal transverse Mercator projection coordinate system; the preset geographic package is GeoPackage format; and the cloud-optimized geotagged image file format is cloud-optimized GeoTIFF format.

Images