A method and system for quantifiable geomorphic evidence screening of water-bearing minerals on the surface of mars
By combining deep learning models with Mars orbiter images and mineral data, quantitative coupling of water-bearing minerals and landform features on the Martian surface has been achieved. This solves the problems of strong subjectivity in manual interpretation and difficulty in quantitative characterization in existing technologies, and supports in-depth research on Martian geological evolution and potential signs of life.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- TONGJI UNIV
- Filing Date
- 2026-04-28
- Publication Date
- 2026-07-14
Smart Images
Figure CN122391783A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of Mars surface data processing technology, and in particular to a method and system for screening quantifiable geomorphological evidence of water-bearing minerals on the Martian surface. Background Technology
[0002] Mars is one of the most Earth-like planets in the solar system. Its surface preserves rich records of ancient water activity and may contain signs of ancient life, making it a key target for deep space exploration and planetary science research. Evidence of water activity on the Martian surface is mainly found in surface geology, mineral composition, and landforms. Among these, Martian water-bearing minerals record paleoenvironmental information related to water, such as river, lake, or ocean sedimentary environments, providing direct indications of early Martian water environments. Studying Martian water-bearing minerals not only helps reveal the geological evolution of Mars but also provides important clues for understanding its climate change patterns and exploring potential signs of life.
[0003] Significant correlations exist between hydrous minerals and geomorphic units on the Martian surface. At a local scale, equatorial layered deposits (ELDs) are widely distributed in the Martian equatorial region, typically exhibiting repeated, superimposed bedding structures and rich erosive landforms, such as yardangs and other typical aeolian erosion patterns. In the Tyrrhena Terra region of southern Mars, hydrous silicates are mainly found in impact crater-related geomorphic units, including ejecta, crater walls, crater rims, and central peaks. At a global scale, statistical analyses of hydrous minerals and geomorphic units worldwide have shown that over 50% of hydrous minerals are associated with impact crater landforms, primarily concentrated in ejecta sheets, crater walls, crater rims, and central peaks. Furthermore, although representing a relatively smaller proportion, hydrous minerals are also commonly found in alluvial fans, deltas, and other sedimentary landforms.
[0004] In planetary geology research, geomorphic evidence refers to clues that can be used to infer the existence of specific geological processes or environmental conditions through the morphological features of a planet's surface. Traditional geomorphic evidence interpretation processes usually rely on experts to manually interpret special landforms in local areas and infer the possible distribution of water ice or ancient water flow activities in the study area.
[0005] However, this process still has significant shortcomings: First, expert interpretation is highly subjective and time-consuming, making it difficult to meet the needs of large-scale, high-efficiency research. Secondly, existing methods are insufficient for quantitatively characterizing the relationship between landforms and water flow or the presence of water ice, thus limiting a deeper understanding of the relevant mechanisms. Summary of the Invention
[0006] The purpose of this invention is to overcome the shortcomings of the existing technology, such as the strong subjectivity of expert manual interpretation and the difficulty in quantitatively characterizing the relationship between landforms and water flow activities or the presence of water ice, and to provide a quantifiable landform evidence screening method and system for water-bearing minerals on the Martian surface.
[0007] The objective of this invention can be achieved through the following technical solutions: A method for screening quantifiable geomorphic evidence of water-bearing minerals on the Martian surface, comprising: Acquire Mars orbiter images of the target study area and Martian surface mineral mapping data of the target study area; The Mars orbiter images are input into a pre-trained deep learning-based terrain interpretation model to identify terrain features and extract terrain characteristics. A water-bearing mineral dataset was constructed using Martian surface mineral mapping data to train a pre-built water-bearing mineral detection model. The trained water-bearing mineral detection model was then used to predict the mineralization pseudoprobability at each point within the target study area based on the Martian orbiter images, thus obtaining the distribution results of water-bearing minerals. Correlation analysis was performed based on the geomorphic features and the distribution results of the hydrous minerals to obtain quantitative coupling results between hydrous minerals and specific geomorphic features within the target study area.
[0008] Furthermore, the processing procedure of the terrain interpretation model includes: The Mars orbiter images are processed into grids, and the trained terrain interpretation model is used to identify terrain features and extract terrain characteristics from local image blocks in each grid.
[0009] Furthermore, the process of acquiring Martian surface mineral mapping data corresponding to the target study area includes: A full Mars pyroxene distribution map, a full Mars water-bearing mineral point database, and a full Mars water-bearing mineral raster database were obtained, respectively. The data were then cropped according to the range of the target study area and resampled to the same resolution to obtain the Martian surface mineral mapping data corresponding to the target study area.
[0010] Furthermore, the scene-level labels of the data in the hydrous mineral dataset are divided into hydrous minerals and non-hydrous minerals; For data on hydrous mineral scenes, firstly, hydrous mineral points in the target study area are selected, and a buffer zone is established with each hydrous mineral point as the center. Then, hydrous mineral raster data is extracted in each buffer zone as candidate hydrous mineral samples, and the data of the candidate hydrous mineral samples is recorded as n. For data on non-aqueous mineral scenarios, firstly, aqueous mineral points in the target study area are selected, and a buffer zone is established with each aqueous mineral point as the center. Then, Martian pyroxene is selected from outside each buffer zone as candidate non-aqueous mineral samples.
[0011] Furthermore, the selection of Martian pyroxenes from outside each buffer zone as candidate non-hydrous mineral samples includes: Sort the Martian pyroxenes outside each buffer according to their abundance values, and select the top n Martian pyroxenes, which are the same number as the candidate hydrous mineral samples, as non-hydrous mineral samples.
[0012] Furthermore, a correlation analysis is performed based on the geomorphic features and the distribution results of the hydrous minerals, including: The Pearson correlation coefficient is used to calculate the correlation between the distribution of the landform features and the distribution of the hydrous minerals, thereby obtaining the quantitative coupling results between hydrous minerals and specific landforms in the target study area.
[0013] Furthermore, the method also includes: if the relative frequency of a certain type of landform is lower than a preset frequency threshold, then the landform is regarded as a rare category and removed.
[0014] Furthermore, the deep learning-based terrain interpretation model is the TinyViT model, which is trained on a publicly available planetary terrain classification dataset.
[0015] Furthermore, the identified landforms include mixed landforms, small impact crater landforms, smooth landforms, and textured landforms.
[0016] The present invention also provides a quantifiable geomorphological evidence screening system for water-bearing minerals on the surface of Mars, including a memory and a processor, wherein the memory stores a computer program, and the processor calls the computer program to execute the steps of the method described above.
[0017] Compared with the prior art, the present invention has the following advantages: (1) This invention constructs a deep learning-driven framework for quantifying geomorphic evidence, transforming the traditional qualitative interpretation method that relies on manual interpretation into a calculable, verifiable, and reproducible quantitative analysis system. This framework establishes a geomorphic interpretation model and a water-bearing mineral detection model. The geomorphic interpretation model identifies geomorphic features and extracts geomorphic characteristics based on Mars orbiter images of the target study area. The water-bearing mineral detection model is trained based on mineral mapping data on the Martian surface to first obtain the distribution results of water-bearing minerals based on Mars orbiter images. Finally, by combining correlation modeling and uncertainty estimation methods, the system reveals and quantifies the coupling relationship between Martian surface geomorphic units and the distribution of water-bearing minerals, providing important scientific evidence for a deeper understanding of the hydrological processes and water-bearing mineral formation mechanisms on the Martian surface.
[0018] (2) This invention makes full use of existing Martian mineral spatial distribution data and Martian high-resolution optical image data to construct a cross-scale, multi-source data fusion training and verification system, realizes joint modeling and predictive analysis of specific geomorphic units and water-bearing minerals, and provides new scientific basis and technical support for understanding the evolution of the Martian paleoenvironment, as well as future Martian landing site selection and resource exploration. Attached Figure Description
[0019] Figure 1 This is a flowchart illustrating a method for screening quantifiable geomorphological evidence of water-bearing minerals on the surface of Mars, provided in an embodiment of the present invention. Figure 2 This is a flowchart of a geomorphological mineral correlation analysis provided in an embodiment of the present invention; Figure 3 This is a geomorphological mineral correlation analysis result diagram provided in an embodiment of the present invention. Detailed Implementation
[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0021] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.
[0022] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.
[0023] Example 1 like Figure 1 As shown, this embodiment provides a method for screening quantifiable geomorphic evidence of water-bearing minerals on the Martian surface, including: S1: Acquire Mars orbiter images of the target study area and Martian surface mineral mapping data of the target study area; S2: Input the Mars orbiter images into a pre-trained deep learning-based terrain interpretation model to identify terrain features and extract terrain characteristics; S3: Construct a water-bearing mineral dataset using Martian surface mineral mapping data to train a pre-built water-bearing mineral detection model. Using the trained water-bearing mineral detection model, predict the mineralization pseudo-probability at each point in the target study area based on Martian orbiter images, and obtain the distribution results of water-bearing minerals. S4: Based on the geomorphological features and the distribution results of hydrous minerals, a correlation analysis is conducted to obtain the quantitative coupling results between hydrous minerals and specific geomorphological features within the target study area.
[0024] Specifically, in step S1, the process of acquiring Martian surface mineral mapping data corresponding to the target study area includes: We acquired a full Mars pyroxene distribution map, a full Mars point database of water-bearing minerals, and a full Mars raster database of water-bearing minerals, respectively. We then cropped the data according to the target study area and resampled it to the same resolution to obtain the Martian surface mineral mapping data corresponding to the target study area.
[0025] In step S2, the processing of the terrain interpretation model includes: The Mars orbiter images were processed into grids, and the trained terrain interpretation model was used to identify terrain features and extract terrain characteristics from local image blocks in each grid.
[0026] Optionally, the deep learning-based terrain interpretation model is the TinyViT model, which is trained on a publicly available planetary terrain classification dataset.
[0027] The specific steps are as follows: A lightweight visual Transformer model (TinyViT) is trained based on a publicly available planetary landform classification dataset. To meet the processing needs of large-scale, high-resolution optical remote sensing imagery, the original image is first divided into regular grids, each with a spatial size of 200×200 pixels. Then, the trained TinyViT model is used to automatically identify landform categories in local image patches within each grid, and simultaneously extracts the corresponding semantic features of the landforms.
[0028] In step S3, the scene-level labels of the data in the hydrous mineral dataset are divided into hydrous minerals and non-hydrous minerals; For data on hydrous mineral scenes, firstly, hydrous mineral points in the target study area are selected, and a buffer zone is established with each hydrous mineral point as the center. Then, hydrous mineral raster data is extracted in each buffer zone as candidate hydrous mineral samples, and the data of the candidate hydrous mineral samples is recorded as n. For data on non-aqueous mineral scenarios, we first screen out aqueous mineral points in the target study area and establish a buffer zone centered on each aqueous mineral point. Then, we select Martian pyroxene from outside each buffer zone as candidate non-aqueous mineral samples.
[0029] The specific steps of step S3 are as follows: S301: A dataset of hydrous minerals was constructed. The image size was set to 200×200, and the scene-level labels were divided into two categories: hydrous minerals and non-hydrous minerals. S302: For the water-bearing mineral scenario, we first screened all water-bearing mineral points obtained based on the CRISM target detection mode within the study area. Considering the small sample size, we performed data augmentation by establishing a buffer zone centered on each water-bearing mineral point. Low-resolution water-bearing mineral raster data was extracted within each buffer as candidate water-bearing mineral samples, with the sample size denoted as n. S303: For non-aqueous mineral scenarios, based on the construction of a aqueous mineral buffer zone, Martian pyroxene is randomly selected from outside the buffer zone as candidate non-aqueous mineral samples. Specifically, the abundance values of pyroxene within the candidate regions are sorted, and then the first n samples with the same number as the candidate aqueous mineral samples are selected as non-aqueous mineral samples; S304: A water-bearing mineral detection model is trained based on a large-scale water-bearing mineral dataset, and the model is used to infer mineralization pseudoprobabilities from Mars orbiter optical images.
[0030] In step S4, a correlation analysis is performed based on the geomorphological features and the distribution results of water-bearing minerals, including: The Pearson correlation coefficient was used to calculate the correlation between the distribution of geomorphic features and the distribution of hydrous minerals, so as to obtain the quantitative coupling results between hydrous minerals and specific geomorphic features in the target study area.
[0031] Preferably, the method further includes: if the relative frequency of a certain type of landform is lower than a preset frequency threshold, then the landform is regarded as a rare category and removed.
[0032] The specific steps of step S4 include: S401: The Pearson correlation coefficient is used to quantitatively assess the correlation between geomorphic features and hydrous minerals. The calculation formula is as follows: in, yes and The covariance. yes variance yes The variance.
[0033] S402: Considering that insufficient sample size may lead to unstable correlation estimation results, when the relative frequency of a certain landform is less than 0.5%, it is regarded as a rare category and removed. S403: The correlation coefficient is resampled using the bootstrap method to construct its empirical sampling distribution, and the confidence interval is calculated accordingly; S404: Introduce multiple comparison correction during the significance level (p > 0.05) test to control false positive inflation caused by multiple tests.
[0034] The following is a specific example of the above solution: Step S1: Acquire high-resolution Mars orbiter images and Martian mineral data corresponding to the target study area; Step S101: Based on the scope of the study area, in NASA's Planetary Data System (PDS) (Source: https: / / astrogeology.usgs.gov / search / map / mars-mro-ctx-global-mosaic-murray-lab-v1) Query and download all CTX images within the study area, and crop them according to the boundaries of the Marswas Valley region to generate a high-resolution optical image dataset.
[0035] Step S102: Download the full Mars pyroxene distribution map and the full Mars water-bearing mineral point database (source: https: / / ode.rsl.wustl.edu / ), and crop the dataset according to the scope of the study area.
[0036] Step S103: Download the full Mars water-bearing mineral raster database (source: https: / / www.ias.u-psud.fr / moccas / ) and crop it according to the scope of the study area to generate a dataset.
[0037] Step S104: Preprocess the acquired Martian pyroxene distribution map and hydrous mineral raster database, i.e., resample to the same resolution.
[0038] Step S2: Automatically identify landforms and extract landform features using a deep learning-based landform interpretation model; Step S201: Train the TinyViT model based on the DoMars16k dataset (source: https: / / zenodo.org / records / 4291940). Large-scale high-resolution optical image data is processed into a grid with a grid size of 200×200. The trained model is then used to automatically identify terrain features and extract terrain characteristics from the corresponding images within each grid.
[0039] Step S3: Predict the false probability of mineralization in the region using a hydrous mineral detection model; Step S301: Construct a hydrous mineral dataset based on the mineral dataset within the study area. Scene-level labels are divided into two categories: hydrous minerals and non-hydrous minerals. For hydrous mineral scenarios, all hydrous mineral points obtained based on the CRISM target detection mode within the study area are selected. Considering the small sample size, data augmentation is performed, and a buffer is established centered on each hydrous mineral point. Low-resolution hydrous mineral raster data is extracted within each buffer as candidate hydrous mineral samples, with the sample size denoted as n. For non-hydrous mineral scenarios, based on the constructed hydrous mineral buffer, samples from outside the buffer are sorted according to pyroxene abundance values, and the top n samples with the same number as the candidate hydrous mineral samples are selected as non-hydrous mineral samples.
[0040] Step S302: Train a water-bearing mineral detection model based on the constructed large-scale water-bearing mineral dataset, and use the model to infer the mineralization pseudo-probability of the corresponding high-resolution optical image in each grid.
[0041] Step S4: Use correlation analysis to screen landforms and establish quantitative coupling patterns between Martian water-bearing minerals and specific landform units within the region; Step S401: See Figure 2First, the Pearson correlation coefficient was used to quantitatively assess the correlation between geomorphic features and hydrous minerals. The calculation formula is as follows: in yes and The covariance. yes variance yes The variance.
[0042] Step S402: Considering that insufficient sample size may lead to unstable correlation estimation results, when the relative frequency of a certain landform is less than 0.5%, it is regarded as a rare category and removed. Step S403: Use the bootstrap method to perform uniform sampling with replacement on the data to construct an empirical sampling distribution of the correlation coefficient, and calculate its confidence interval based on the distribution; Step S404: During the significance level (p > 0.05) test, multiple comparison correction is introduced to control the inflation of the false positive rate caused by multiple tests. Specifically, the False Discovery Rate (FDR) correction method is used. By controlling the expected proportion of false positive results among all results judged as significant, the risk of false discovery caused by multiple tests is effectively reduced.
[0043] The experiment is as follows: 1. Experimental Data High-resolution CTX images of the Maws Valley region and their corresponding Martian surface mineral mapping data were used as experimental data to test the geomorphic evidence screening method of this invention.
[0044] Experimental results according to Figure 3 As a result, within the Maws Valley region, textured landforms (tex) and mixed landforms showed a significant positive correlation with the distribution of water-bearing minerals. Both types of landforms were formed under complex geological processes, and their evolution is likely closely related to ancient water flow activities. This discovery not only deepens our understanding of the Martian water environment and its geological evolution but also provides important scientific evidence for future Mars exploration missions, especially in the analysis of Martian habitability. Conversely, smooth landforms (smooth landforms) and small impact crater landforms (sfx) showed a significant negative correlation.
[0045] According to the results in Table 1, in the Maws Valley region, both textured landforms (tex) and mixed landforms showed a significant positive correlation with hydrous minerals in terms of spatial distribution, indicating that the spatial distribution patterns of these two types of landforms are highly similar to the spatial patterns of hydrous mineral enrichment areas. In contrast, smooth landforms (smo) and small impact crater landforms (sfx) showed a negative correlation with hydrous minerals. Notably, the Moran's I for mixed landforms in this region reached 0.338, significantly higher than that for textured landforms (tex), indicating a closer spatial coupling between mixed landforms and hydrous minerals. The formula for calculating the Moran's I is as follows: in, , This represents the total number of spatial units. and They represent the first The first spatial unit and the first The attribute values of each spatial unit. The mean of all spatial unit attribute values. This represents the spatial weight value.
[0046] Table 1. Results of spatial correlation analysis of hydrous minerals and geomorphic units based on the bivariate Moran index. This embodiment also provides a system for screening quantifiable geomorphic evidence of water-bearing minerals on the Martian surface, including a memory and a processor. The memory stores a computer program, and the processor calls the computer program to execute the steps of the above-described method for screening quantifiable geomorphic evidence of water-bearing minerals on the Martian surface.
[0047] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.
Claims
1. A method for screening quantifiable geomorphic evidence of water-bearing minerals on the Martian surface, characterized in that, include: Acquire Mars orbiter images of the target study area and Martian surface mineral mapping data of the target study area; The Mars orbiter images are input into a pre-trained deep learning-based terrain interpretation model to identify terrain features and extract terrain characteristics. A water-bearing mineral dataset was constructed using Martian surface mineral mapping data to train a pre-built water-bearing mineral detection model. The trained water-bearing mineral detection model was then used to predict the mineralization pseudoprobability at each point within the target study area based on the Martian orbiter images, thus obtaining the distribution results of water-bearing minerals. Correlation analysis was performed based on the geomorphic features and the distribution results of the hydrous minerals to obtain quantitative coupling results between hydrous minerals and specific geomorphic features within the target study area.
2. The method for screening quantifiable geomorphic evidence of water-bearing minerals on the Martian surface according to claim 1, characterized in that, The processing steps of the terrain interpretation model include: The Mars orbiter images are processed into grids, and the trained terrain interpretation model is used to identify terrain features and extract terrain characteristics from local image blocks in each grid.
3. The method for screening quantifiable geomorphic evidence of water-bearing minerals on the Martian surface according to claim 1, characterized in that, The process of acquiring Martian surface mineral mapping data corresponding to the target study area includes: A full Mars pyroxene distribution map, a full Mars water-bearing mineral point database, and a full Mars water-bearing mineral raster database were obtained, respectively. The data were then cropped according to the range of the target study area and resampled to the same resolution to obtain the Martian surface mineral mapping data corresponding to the target study area.
4. The method for screening quantifiable geomorphic evidence of water-bearing minerals on the Martian surface according to claim 3, characterized in that, The scene-level labels of the data in the hydrous mineral dataset are divided into hydrous minerals and non-hydrous minerals; For data on hydrous mineral scenes, firstly, hydrous mineral points in the target study area are selected, and a buffer zone is established with each hydrous mineral point as the center. Then, hydrous mineral raster data is extracted in each buffer zone as candidate hydrous mineral samples, and the data of the candidate hydrous mineral samples is recorded as n. For data on non-aqueous mineral scenarios, firstly, aqueous mineral points in the target study area are selected, and a buffer zone is established with each aqueous mineral point as the center. Then, Martian pyroxene is selected from outside each buffer zone as candidate non-aqueous mineral samples.
5. The method for screening quantifiable geomorphic evidence of water-bearing minerals on the Martian surface according to claim 4, characterized in that, The selection of Martian pyroxenes from outside each buffer zone as candidate anhydrous mineral samples includes: Sort the Martian pyroxenes outside each buffer according to their abundance values, and select the top n Martian pyroxenes, which are the same number as the candidate hydrous mineral samples, as non-hydrous mineral samples.
6. The method for screening quantifiable geomorphic evidence of water-bearing minerals on the Martian surface according to claim 1, characterized in that, A correlation analysis was performed based on the geomorphic features and the distribution results of the hydrous minerals, including: The Pearson correlation coefficient is used to calculate the correlation between the distribution of the landform features and the distribution of the hydrous minerals, thereby obtaining the quantitative coupling results between hydrous minerals and specific landforms in the target study area.
7. The method for screening quantifiable geomorphic evidence of water-bearing minerals on the Martian surface according to claim 1, characterized in that, The method further includes: if the relative frequency of a certain type of landform is lower than a preset frequency threshold, then the landform is regarded as a rare category and removed.
8. The method for screening quantifiable geomorphic evidence of water-bearing minerals on the Martian surface according to claim 1, characterized in that, The deep learning-based terrain interpretation model is the TinyViT model, which is trained on a publicly available planetary terrain classification dataset.
9. The method for screening quantifiable geomorphic evidence of water-bearing minerals on the Martian surface according to claim 1, characterized in that, The identified landforms include mixed landforms, small impact crater landforms, smooth landforms, and textured landforms.
10. A quantifiable geomorphic evidence screening system for water-bearing minerals on the Martian surface, characterized in that, It includes a memory and a processor, the memory storing a computer program, the processor invoking the computer program to perform the steps of the method as described in any one of claims 1 to 9.