Hyperspectral hydrous mineral detection method and system based on Mars orbiter

Abstract

The invention discloses a hydrous mineral detection method and system based on the hyperspectrum of a Mars orbiter, and relates to the technical field of hyperspectral data processing and analys.The method comprises the steps that a distribution area of hydrous minerals is obtained based on Mars hyperspectral data; performing automatic end member extraction on the distribution area to obtain an image end member; and performing spectral feature inspection on the image end members, and performing unmixing based on an image end member library to obtain a hydrous mineral abundance map. According to the method, the characteristics of the water-containing minerals can be automatically identified, the mineral distribution range can be accurately positioned, meanwhile, the de-mixing end member library can be automatically constructed, the end members can be checked, and manual participation is reduced; the end member library is automatically constructed from the image based on the qualitative recognition result, and through overlay analysis and comparison with other unmixing end member construction methods, it is shown that the method has high credibility in the aspect of reflecting the distribution of the hydrous minerals and the abundance of the minerals.

Description

Technical Field

[0001] The present invention relates to the technical field of hyperspectral data processing and analysis, and in particular to a method and system for detecting water-containing minerals based on Mars orbiter hyperspectral data. Background Art

[0002] Mars orbiter hyperspectral data has rich spectral information and can provide detailed data on the mineral composition and distribution of the Martian surface, especially playing an important role in the detection of hydrous minerals. With the gradual deepening of the application of Mars orbiter hyperspectral data in the detection of hydrous minerals, refined and highly reliable hyperspectral data processing technology has become one of the current research hotspots in the detection of hydrous minerals on Mars. Qualitative and quantitative analysis combined with Mars hyperspectral data can not only provide distribution information of hydrous minerals, but also accurately estimate their abundance, providing an important mineralogical basis for studying the geological activities and evolution processes of Mars.

[0003] At present, the methods for detecting water-bearing minerals on Mars have the following main shortcomings: First, the degree of automation is low. Existing methods usually rely on manual construction of end-member spectral libraries and visual inspection of extracted image end-members, and qualitative and quantitative detection are usually carried out separately, which cannot be fully automated; this process with a lot of manual intervention increases the complexity of the operation and limits its application in large-scale hyperspectral data analysis; second, the credibility of the detection results is low. Existing methods usually construct unmixed spectral libraries based on actual spectra measured in Earth laboratories, which cannot fully reflect the spectral characteristics of real Martian surface minerals. This difference leads to a large deviation between the unmixing results and the actual mineral composition of Mars during quantitative inversion; in addition, the Earth spectral library often cannot cover the complex mineral combinations and environmental conditions on the Martian surface. This step affects the accuracy and credibility of the detection results; finally, it is difficult to detect the distribution of low-abundance hydrous minerals. Since the spectral characteristics of low-abundance hydrous minerals often manifest as weak absorption characteristics and are easily masked by other background spectra or noise signals, traditional methods often have difficulty in accurately separating them during spectral unmixing and feature identification; secondly, spectral signals under low signal-to-noise ratio conditions will further aggravate the difficulty of feature identification and lead to misjudgment. Existing methods usually lack special optimization design for weak spectral features and cannot effectively suppress background interference or adapt to spectral changes in complex environments; therefore, there is an urgent need to design a qualitative and quantitative detection method for hydrous minerals based on Mars orbiter hyperspectral to improve the degree of automation, enhance the credibility of detection results, and accurately obtain the spatial distribution and abundance results of hydrous minerals. Summary of the Invention

[0004] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and in the abstract and title of the present invention to avoid obscuring the purpose of this section, the abstract and the title of the invention, and such simplifications or omissions should not be used to limit the scope of the present invention.

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the present invention provides a method and system for detecting water-containing minerals based on Mars orbiter hyperspectral, which can solve the problems mentioned in the background technology.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In the first aspect, an embodiment of the present invention provides a method for detecting hydrated minerals based on Mars orbiter hyperspectral data, which includes obtaining the distribution area of ​​hydrated minerals based on Mars hyperspectral data; performing automatic end-member extraction on the distribution area to obtain image end-members; performing spectral feature inspection on the image end-members, and unmixing based on the image end-member library to obtain hydrated mineral abundance mapping.

[0009] As a preferred solution of the method for detecting hydrated minerals based on Mars orbiter hyperspectral data described in the present invention, the method of obtaining the distribution area of ​​hydrated minerals based on Mars hyperspectral data refers to adopting a target detection method for Mars hyperspectral data, using the spectra of hydrated minerals in the reference spectral library as a guide, and obtaining high-confidence distribution areas of various types of hydrated minerals; the method of adopting a target detection method for Mars hyperspectral data includes: using a hierarchical minimum output energy constraint method, using the spectra of hydrated minerals in the reference spectral library as a guide, and performing target detection on Mars hyperspectral data.

[0010] As a preferred solution of the method for detecting hydrated minerals based on Mars orbiter hyperspectral analysis according to the present invention, the automated end-member extraction of the distribution area includes: presetting the number of end-members N for the high-confidence distribution areas of each type of hydrated minerals, using vertex component analysis as the end-member extraction method, and obtaining image end-members based on the set number of end-members N.

[0011] As a preferred solution of the method for detecting hydrated minerals based on Mars orbiter hyperspectral data described in the present invention, the hydrated mineral abundance mapping includes the following steps: an end-member inspection method based on spectral angle and spectral absorption feature inspection is used to eliminate erroneously extracted noise and background pixels in the image end-members to obtain high-confidence hydrated mineral end-members in the image; an image unmixing end-member library is automatically constructed to unmix the hyperspectral data to obtain various hydrated mineral abundance results and the total abundance of hydrated minerals on Mars.

[0012] As a preferred solution of the method for detecting hydrated minerals based on Mars orbiter hyperspectral analysis described in the present invention, the end-member verification method based on spectral angle includes: using the spectral angle matching method to perform global matching on the overall spectral shape, establishing a buffer zone based on the absorption center position and absorption width of various types of hydrated minerals in the spectral library, and constraining the local spectral absorption characteristics. Image end-members that meet both global matching and local constraints are considered to be high-confidence hydrated mineral end-members.

[0013] As a preferred solution of the method for detecting hydrated minerals based on the Mars orbiter hyperspectral method described in the present invention, the end member inspection method based on spectral absorption feature inspection includes: constructing absorption features based on the spectral library, defining the absorption interval to be inspected and the absorption center of the spectral library, and constructing a buffer range for the absorption interval to be inspected considering the spectral variability; locating the absorption left endpoint, absorption right endpoint and absorption center of the image end member according to the gradient method, and the image end member is considered to be correctly extracted if and only if the three positions of all absorption intervals fall within the allowable range of the reference spectrum, otherwise the corresponding image end member is eliminated.

[0014] As a preferred solution of the method for detecting hydrous minerals based on Mars orbiter hyperspectral detection according to the present invention, the hydrous mineral abundance mapping includes using different colors to represent the abundance range of hydrous minerals on Mars.

[0015] Secondly, in order to further solve the security problems existing in the processing and analysis of hyperspectral data, the present invention provides a water-containing mineral detection system based on the Mars orbiter hyperspectral data, which includes: an area acquisition module for acquiring the distribution area of ​​water-containing minerals based on Mars hyperspectral data; an end member extraction module for automatically extracting end members from the distribution area to obtain image end members; an abundance mapping module for performing spectral feature inspection on the image end members, and performing unmixing based on the image end member library to obtain a water-containing mineral abundance map.

[0016] The beneficial effects of the present invention are as follows: the present invention proposes an integrated qualitative and quantitative detection method for hydrated minerals based on the Mars orbiter hyperspectral image, which can automatically identify the characteristics of hydrated minerals and accurately locate the distribution range of minerals. At the same time, it can automatically construct an unmixed end-member library and test the end-members, reducing manual participation. It can be flexibly applied to data of different resolutions, thereby improving the degree of automation of the Mars exploration mission. Based on the qualitative recognition results, the end-member library is automatically constructed from the image. Through overlay analysis and comparison with other unmixed end-member construction methods, it is shown that the present invention has a high degree of credibility in reflecting the distribution and mineral abundance of hydrated minerals. By performing automated end-member extraction in the detected hydrated mineral distribution area, image end-members can be effectively extracted in the low-abundance hydrated mineral distribution area. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort. Among them:

[0018] Figure 1 This is an overall flow chart of the method for detecting water-containing minerals based on Mars orbiter hyperspectral in Example 1.

[0019] Figure 2 This is a schematic diagram of the binary detection results of various types of water-containing minerals using Mars CRISM data as an example in Example 3.

[0020] Figure 3 This is a schematic diagram of the end member extraction results using Mars CRISM data as an example in Example 3.

[0021] Figure 4 This is a schematic diagram of the results of mapping the abundance of water-containing minerals on Mars in Example 3.

[0022] Figure 5 This is a schematic diagram of dividing the hydrous mineral regions with strong, medium and weak characteristics and the background region based on the spectral index results in Example 3. DETAILED DESCRIPTION

[0023] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0024] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0025] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0026] Example 1

[0027] Reference Figure 1 , which is the first embodiment of the present invention, provides a method for detecting water-containing minerals based on Mars orbiter hyperspectral.

[0028] The existing methods for detecting hydrated minerals on Mars have the following main problems: First, the degree of automation is low. Existing methods usually rely on manual construction of endmember spectral libraries and visual inspection of extracted image endmembers, and qualitative and quantitative detection are usually carried out separately, which cannot be fully automated. This process with a lot of manual intervention increases the complexity of the operation and limits its application in large-scale hyperspectral data analysis. Second, the detection results are of low credibility. Existing methods usually construct unmixed spectral libraries based on actual spectra measured in Earth laboratories, which cannot fully reflect the spectral characteristics of real Martian surface minerals. This difference leads to a large discrepancy between the unmixing results and the actual mineral composition of Mars during quantitative inversion. deviation; in addition, the Earth's spectral library often cannot cover the complex mineral combinations and environmental conditions on the surface of Mars, further affecting the accuracy and credibility of the detection results; finally, it is difficult to detect the distribution of low-abundance hydrous minerals. Since the spectral characteristics of low-abundance hydrous minerals often show weak absorption characteristics and are easily masked by other background spectra or noise signals, traditional methods often have difficulty in accurately separating them during spectral unmixing feature identification; secondly, spectral signals under low signal-to-noise ratio conditions will further aggravate the difficulty of feature identification and lead to misjudgment. Existing methods usually lack special optimization design for weak spectral features and cannot effectively suppress background interference or adapt to spectral changes in complex environments.

[0029] This application provides an effective solution to the above-mentioned problems. Next, we will combine multiple embodiments to explain in detail how to implement the method for detecting water-containing minerals based on the Mars orbiter hyperspectral.

[0030] Figure 1 The overall flow chart of the method for detecting hydrous minerals based on Mars orbiter hyperspectral imaging is shown, including:

[0031] S1: Obtain the distribution area of ​​hydrated minerals based on Mars hyperspectral data.

[0032] In an embodiment of the present application, obtaining the distribution area of ​​hydrated minerals based on Mars hyperspectral data refers to using a target detection method for Mars hyperspectral data, using the spectra of hydrated minerals in a reference spectral library as a guide, to obtain high-confidence distribution areas of various types of hydrated minerals.

[0033] In an embodiment of the present application, a target detection method for Mars hyperspectral data includes: using a hierarchical constrained energy minimization (hCEM) method to perform target detection on Mars hyperspectral data using the spectra of water-containing minerals in a reference spectral library as a guide.

[0034] For example, in the embodiment of the present application, the specific formula for obtaining the high confidence distribution area of ​​each type of hydrous mineral is as follows:

[0035]

[0036] Where d is the spectrum of hydrous minerals in the reference spectrum library; X k is the residual background data obtained after the original hyperspectral X is processed by k layers; R is the background covariance matrix; y k It is the hyperspectral target detection result after the k-th layer filter output.

[0037] In an optional embodiment, when using a target detection method for Mars hyperspectral data, multiple algorithms can be combined to adapt to different scenario requirements; for example, in a traditional spectral matching scenario, an adaptive matched filter can be used to detect hydrous minerals, wherein the matching degree between the pixel to be tested and the reference spectrum is calculated, and the background covariance matrix is ​​combined to suppress noise interference, which is suitable for areas with a high signal-to-noise ratio; in addition, the orthogonal subspace projection method can also be used for target detection, by projecting the background signal into a subspace orthogonal to the target spectrum to further separate the target and background signals, which is particularly suitable for scenarios with complex backgrounds but significant target spectral features; in complex dynamic environments, a hybrid detection framework based on deep learning can also be introduced, such as designing a three-dimensional convolutional neural network to jointly extract spectral and spatial features, optimizing target detection results through end-to-end training, or using a generative adversarial network to generate potential spectral variants of hydrous minerals to enhance the model's robustness to spectral variability; for multi-temporal data, the spatiotemporal attention mechanism can also be combined to adjust the weights of different bands and regions to improve adaptability to spectral shifts caused by seasonal changes on Mars.

[0038] It should be noted that the present invention solves the key technical difficulties in identifying hydrated minerals in Mars hyperspectral data by adopting a layered minimum output energy constraint method. Among them, since the traditional single-layer energy constraint method is often affected by background interference when dealing with the complex surface environment of Mars, resulting in inaccurate identification results of hydrated minerals, the present invention uses a layered processing mechanism to enable the detection of each layer to gradually eliminate background interference, effectively overcoming the pollution of spectral signals by complex environmental factors such as dust cover, volcanic ash and sandstorms on the surface of Mars. Especially in areas with weak absorption characteristics of hydrated minerals, the layered minimum output energy constraint method can enhance the signal-to-noise ratio and highlight the characteristic signals of hydrated minerals, thereby greatly improving the sensitivity and accuracy of detection. Compared with traditional methods, the layered minimum output energy constraint method can effectively reduce the misjudgment rate caused by factors such as changes in lighting conditions and atmospheric interference, so that accurate identification can be achieved even in typical low-abundance hydrated mineral distribution areas on Mars.

[0039] S2: Automatically extract endmembers from the distribution area to obtain image endmembers.

[0040] In an embodiment of the present application, automated endmember extraction of the distribution area includes: presetting the number of endmembers N for the high-confidence distribution area of ​​each type of water-containing mineral, using Vertex Component Analysis (VCA) as the endmember extraction method, and obtaining image endmembers based on the set number of endmembers N.

[0041] In an optional embodiment, automated endmember extraction of the distribution area can also be achieved by using a hybrid strategy based on traditional optimization algorithms or combined with spatial constraints. For example, in a complex spectral mixing scenario, the hyperspectral data can be decomposed into the product of the endmember matrix and the abundance matrix through a non-negative matrix decomposition algorithm, and the purity and interpretability of the endmembers can be improved by introducing sparsity constraints or manifold learning constraints. In addition, an improved vertex component analysis method can be designed in combination with spatial context information. For example, a morphological reconstruction algorithm can be introduced in the endmember extraction process to constrain the spatial continuity of the endmember through the spectral similarity of spatial neighborhood pixels, thereby avoiding the problem of endmember drift caused by noise interference. For multi-temporal Mars hyperspectral data, a spatiotemporal joint endmember extraction framework can be used to construct an endmember library with temporal consistency by fusing the spectral characteristics and spatial distribution information of different phases. In addition, an adaptive endmember number determination method can also be introduced. For example, based on the virtual dimension estimation technology, the number of endmembers N can be dynamically determined by analyzing the eigenvalue distribution of the data covariance matrix to avoid subjective errors caused by manual pre-setting. Regardless of the method used, it is necessary to ensure that the endmember extraction results can reflect the true spectral characteristics of Martian water-containing minerals. At the same time, the physical rationality of the endmembers must be verified in combination with the spectral library. For example, the endmember screening process can be constrained by the matching degree of the spectral absorption peak position and width.

[0042] It should be noted that the use of vertex component analysis (VCA) for automated endmember extraction effectively addresses the technical issues of low automation and high manual intervention in endmember extraction from Mars hyperspectral data. Traditional endmember extraction methods often rely on expert experience, making it difficult to ensure efficiency and consistency when considering the massive and high-dimensional nature of Mars hyperspectral data. In contrast, the vertex component analysis (VCA) method used in the present invention, based on the geometric convex hull principle, achieves rapid and accurate endmember extraction by automatically calculating the vertex positions in the high-dimensional spectral space. Furthermore, the design of a preset number of endmembers, N, makes the endmember extraction process controllable, avoiding overfitting or underfitting problems. This automated endmember extraction method significantly reduces human resource consumption, improves data processing efficiency and the repeatability of results, and is particularly suitable for processing large-scale, multi-region hyperspectral datasets acquired by Mars orbiters. Compared with traditional methods, the present invention can effectively shorten endmember extraction time while maintaining a high accuracy rate, providing a solid foundation for subsequent precise mapping of hydrous minerals.

[0043] S3: Check the spectral characteristics of the image end members and perform unmixing based on the image end member library to obtain the abundance map of hydrous minerals.

[0044] In the embodiment of the present application, obtaining a hydrous mineral abundance map includes the following steps: removing erroneously extracted noise and background pixels from image endmembers based on an endmember inspection method for checking spectral angles and spectral absorption characteristics, and obtaining high-confidence hydrous mineral endmembers in the image;

[0045] An image unmixing end-element library is automatically constructed to unmix the hyperspectral data to obtain the abundance results of various hydrated minerals and the total abundance of hydrated minerals on Mars.

[0046] In an embodiment of the present application, the end member verification method based on spectral angle includes: using the spectral angle matching (SpectralAngle Mapper, SAM) method to globally match the overall spectral shape, establishing a buffer zone based on the absorption center position and absorption width of various types of hydrated minerals in the spectral library, and constraining the local spectral absorption characteristics. Image end members that meet both global matching and local constraints are considered to be high-confidence hydrated mineral end members.

[0047] For example, in the embodiment of the present application, the specific formula of the end member detection method based on the spectral angle is as follows:

[0048]

[0049] Where SAM is the spectral angle similarity between the image endmember spectrum and the reference spectrum library spectrum; x i with x * The spectra of the end members and the reference spectrum library are extracted for the image respectively, and the threshold is set to remove the image end members that have a large global difference with the reference spectrum library.

[0050] In the embodiment of the present application, the end member inspection method based on the spectral absorption feature inspection includes: constructing the absorption feature according to the spectral library, defining the absorption interval to be inspected [L M ,R M ] and the absorption center C of the spectral library M , and considering the spectral variability, a buffer range is constructed for the absorption interval to be examined

[0051] Locate the absorption left endpoint L of the image end member according to the gradient method IMG , absorb the right endpoint R IMG and absorption center C IMG , the image endmember is considered to be correctly extracted if and only if all three positions of all absorption intervals fall within the allowed range of the reference spectrum, otherwise the corresponding image endmember is eliminated.

[0052] In the embodiment of the present application, the buffer range The specific formula is as follows:

[0053]

[0054] Furthermore, the absorption left endpoint L of the image end member IMG , absorb the right endpoint R IMG and absorption center C IMG The specific formula is as follows:

[0055]

[0056] Where S is the spectral reflectance value of the image end member; bd is the wavelength value of the image end member.

[0057] In an embodiment of the present application, mapping the abundance of hydrous minerals includes using different colors to represent the abundance range of hydrous minerals on Mars.

[0058] In an optional embodiment, the endmember verification method based on spectral angle and spectral absorption feature inspection can select different verification strategies according to data characteristics; for example, when the spectral angle similarity of the image endmember and the matching error with the reference spectral library are large, the spectral reconstruction verification based on the physical model can be preferentially adopted. The theoretical spectral curve of hydrated minerals in the Martian environment is simulated by the Hapke model, and the absorption depth, symmetry and other characteristics of the actual endmember spectrum are compared to eliminate false endmembers caused by illumination or particle effects; if the spectral angle matching is high but the absorption feature position is offset, the dynamic buffer adjustment mechanism is activated to adaptively expand the allowable range of the absorption center position according to the influence of seasonal dust cover on the spectrum of Mars. This hierarchical verification strategy can effectively distinguish between noise interference and real mineral variation, ensuring the reliability of low-abundance hydrated mineral endmembers.

[0059] In an optional embodiment, the end-member inspection method can introduce a machine learning-assisted decision-making mechanism to enhance robustness; for example, when there is a conflict between the spectral angle matching and the absorption feature inspection results, the pre-trained convolutional neural network can be called to perform morphological classification of the end-member spectrum, and by extracting local extreme points, slope changes and other features of the spectral curve, abnormal spectral patterns caused by instrument noise or background aliasing can be identified; at the same time, the spatial distribution consistency of the end-members is evaluated in combination with the random forest algorithm. If an end-member appears isolated in a spatial neighborhood or contradicts a known geological unit, it is marked as a suspicious target; in addition, an active learning framework can be integrated to use manually annotated high-confidence end-members as training samples to optimize the decision boundary of the classification model, thereby improving the inspection accuracy while ensuring the level of automation.

[0060] In an optional embodiment, the end-member verification process can also integrate multi-source data constraints to improve reliability; for example, in the spectral angle matching stage, the laser altimeter data of the Mars orbiter can be combined to exclude abnormal reflectivity pixels caused by terrain shadows; in the absorption feature inspection, the mineral emissivity data of the thermal infrared band is introduced to cross-verify whether the lattice vibration characteristics of the hydrated minerals are consistent with the visible-near infrared spectrum results; for complex mixed pixels, a multi-end-member unmixing-reconstruction cycle verification method can be used, first based on the abundance unmixing of the preliminarily screened end-member library, and then reconstructing the spectrum according to the unmixing residual. If the deviation between the reconstructed spectrum and the original data exceeds the preset threshold, the end-member library is re-optimized and iteratively calculated. This multi-dimensional constraint mechanism can effectively suppress the interference of non-uniform illumination and dust cover on the Martian surface on the end-member verification.

[0061] It should be noted that the present invention proposes a dual end-member verification method, namely, spectral angle and spectral absorption feature inspection, which solves the false positive problem in the identification of Martian hydrated minerals; due to the wide variety of hydrated minerals on the surface of Mars and the high similarity of spectral features, traditional single verification methods are difficult to effectively distinguish different types of hydrated minerals, especially clay minerals and sulfate minerals; the present invention achieves more refined mineral differentiation by combining global spectral shape matching and local absorption feature constraints; by setting an absorption feature buffer, the spectral variability problem is taken into account, so that the algorithm can adapt to the slight spectral changes of minerals on the surface of Mars; this dual verification mechanism effectively reduces the false positive rate and improves the accuracy of end-member extraction. At the same time, the method of locating the absorption feature position based on the gradient method solves the problem that traditional fixed-band detection is easily affected by noise, so that hydrated minerals can be accurately identified even in areas with low signal-to-noise ratio; the automatically constructed image unmixing end-member library further improves the accuracy of abundance mapping and realizes high-precision quantitative analysis of Martian hydrated minerals.

[0062] In summary, the present invention proposes an integrated qualitative and quantitative detection method for hydrated minerals based on the Mars orbiter hyperspectral image, which can automatically identify the characteristics of hydrated minerals and accurately locate the distribution range of minerals. At the same time, it can automatically construct an unmixed end member library and test the end members, reducing manual participation. It can be flexibly applied to data of different resolutions, thereby improving the degree of automation of the Mars exploration mission. Based on the qualitative recognition results, the end member library is automatically constructed from the image. Through overlay analysis and comparison with other unmixed end member construction methods, it is shown that the present invention has a high credibility in reflecting the distribution and mineral abundance of hydrated minerals. By performing automated end member extraction in the detected hydrated mineral distribution area, image end members can be effectively extracted in the low-abundance hydrated mineral distribution area.

[0063] Example 2 is an embodiment of the present invention, which provides a water-containing mineral detection system based on Mars orbiter hyperspectral data, including: an area acquisition module, used to obtain the distribution area of ​​water-containing minerals based on Mars hyperspectral data; an end member extraction module, used to automatically extract end members from the distribution area to obtain image end members; an abundance mapping module, used to perform spectral feature inspection on the image end members, and unmix based on the image end member library to obtain a water-containing mineral abundance map.

[0064] Example 3 is an embodiment of the present invention, which provides a method for detecting water-containing minerals based on Mars orbiter hyperspectral. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0065] This example applies a target detection method to Mars hyperspectral data and inverts the abundance of Martian hydrous minerals based on the map-projected targeted reduced data record (MTRDR) hyperspectral data (frt000098f7) from the Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) onboard the Mars Reconnaissance Orbiter (MRO).

[0066] First, the target detection method of the layered minimum output energy constraint method is used. The spectra of various hydrous minerals in the reference spectral library are used as guide spectra to perform target detection in the Mars hyperspectral data. The distribution areas of various hydrous minerals on Mars are obtained. The binary detection results of various hydrous minerals using the Mars CRISM data as an example are shown in the figure. Figure 2 shown.

[0067] Secondly, for the distribution areas of various hydrated minerals on Mars, since the content of hydrated minerals on Mars is low and the target areas are more dispersed in the qualitative detection results, the preset number of endmembers is appropriately relaxed and set to N = 3. Based on this preset number of endmembers, the vertex component analysis method (VCA) is used to extract endmembers to obtain pure endmembers. The endmember extraction results of the Mars CRISM data are shown as follows: Figure 3 shown.

[0068] Thirdly, an end-member inspection method based on spectral angle and spectral absorption feature inspection is used to eliminate erroneously extracted noise and background pixels, and obtain high-confidence hydrous mineral end-members in the image.

[0069] Among them, the specific method of end-member inspection based on spectral angle and spectral absorption feature inspection is to use the spectral angle matching method to perform global matching on the overall spectral shape, establish a buffer zone according to the absorption center position and absorption width of various types of hydrated minerals in the spectral library, and constrain the local spectral absorption characteristics. The image end-members that meet both global matching and local constraints are considered to be high-confidence hydrated mineral end-members; for global matching based on spectral angle, the spectral angle of the extracted image end-member and the reference spectral library spectrum is calculated in the characteristic absorption band range of Martian hydrated minerals (1.0 microns to 2.7 microns), and 0.08 is used as the threshold to eliminate image end-members with excessively large spectral angles.

[0070] For local constraints based on spectral absorption characteristics, an absorption characteristic model with spectral absorption center position and absorption width is artificially established based on the absorption characteristics of various Martian hydrous minerals in the reference spectral library, as shown in Table 1. The absorption peak position and absorption range of the absorption characteristic model of the image end member are checked, including: first, the absorption characteristics are constructed according to the spectral library, and the absorption range to be checked is defined as [L M ,R M ], the absorption center of the spectral library is C M , and considering the spectral variability, the buffer absorption feature search window is obtained by proportionally expanding the absorption width, and the buffer range is constructed in this interval In the buffer zone Automatically extract the absorption features of the extracted image end members, and locate the absorption left endpoint L of the absorption interval by the position where the gradient increases and decreases IMG , absorb the right endpoint R IMG and absorption center C IMG , the image endmember is considered to be correctly extracted if and only if the three positions of all local absorption intervals fall within the allowed range of the reference spectrum, otherwise the corresponding image endmember is eliminated.

[0071] Table 1 Data table of absorption characteristic models of various Martian water-containing minerals constructed based on the spectral library

[0072] Mineral Type Absorption center position / micron Absorption width / micron Sulfate monohydrate 2.14、2.39 0.17、0.09 alum 1.48、2.17 0.07、0.19 Hydrous ferric sulfate 1.01、2.23 0.40、0.07 Pyrite potassium alum 2.27 0.11 Sulfate polyhydrate 1.04、1.95 0.44、0.19 plaster 1.44、1.94 0.21、0.15 Burnt plaster 1.91 0.14 Kaolinite 1.40、1.92、2.20 0.08、0.09、0.08 Aluminum montmorillonite 1.91、2.21 0.12、0.10 Pearl Mica 1.44、2.02 0.07、0.07 Illite 2.19、2.34 0.08、0.10 Iron montmorillonite 1.08、1.92 0.12、0.10 Magnesium montmorillonite 1.07、1.91 0.22、0.10 magnesite 1.23、2.30 0.63、0.14 siderite 1.92、2.23 0.07、0.24 analcime 1.40、1.90、2.50 0.12、0.19、0.30 hydrous silica 1.45、1.95、2.19 0.14、0.35、0.26 chloride 1.50 0.08 Epidote 1.88、2.32 0.18、0.17 Prehnite 1.49、2.35 0.12、0.18

[0073] Finally, for the high-confidence hydrous mineral endmembers in the obtained images, an unmixing endmember library is automatically constructed to unmix the hyperspectral data, obtaining various hydrous mineral abundance results and the total hydrous mineral abundance. The Hapke model is used to convert the Mars hyperspectral image to be unmixed and the unmixing endmember library from I / F value to single scattering albedo; the least squares-based linear spectral unmixing method is used to obtain the various hydrous mineral abundance results and superimpose them to generate the total abundance of hydrous minerals on Mars, as shown in the following example: Figure 4The following is the result of mapping the abundance of hydrous minerals on Mars using a rainbow gradient color band. Red indicates high abundance of hydrous minerals, purple indicates low abundance of hydrous minerals, and black indicates that the root mean square error (RMSE) of the unmixing result is greater than the threshold of 10. -5 The pixels where unmixing failed.

[0074] In order to further evaluate the effectiveness of the proposed method for qualitative and quantitative detection of hydrous minerals based on Mars orbiter hyperspectral data, the unmixing method based on laboratory spectra and the saliency-based autonomous endmember detection (SAED) method were selected to compare the unmixing of hyperspectral data. Due to the lack of Mars ground truth data for verification, this example is based on the qualitative results of the Mars hydrous mineral browsing product in the CRISM data. The hydrous mineral areas with strong, medium and weak features and the background area are divided according to the index range. Figure 5 As shown in Table 2, the accuracy comparison of each method in the four areas is shown.

[0075] Table 2 Comparison of the accuracy of the proposed method and other methods in various regions

[0076]

[0077] The overall accuracy (OA) of the superposition of the four types of marked areas and the abundance results of hydrous minerals is used as an indicator to evaluate the credibility of the Mars hydrous mineral detection results. The specific formula is as follows:

[0078]

[0079] Among them, C i is the number of correctly classified pixels, that is, the number of correctly classified pixels in category i; N is the total number of categories; M is the number of all pixels in the image. The larger the overall accuracy OA value is, the higher the degree of consistency between the abundance result and the marked area.

[0080] From the comparison results in Table 2, it can be seen that for strong-feature hydrous mineral areas and background areas, all methods can obtain higher accuracy, and the method proposed in the present invention has the best accuracy, which is above 90%; however, for areas with weak hydrous mineral content, compared with unmixing using only the laboratory pure mineral spectrum library, the method proposed in the present invention introduces spectral features as constraints, so it can obtain higher accuracy for areas with weak features; for medium-feature hydrous mineral areas, the accuracy of the method proposed in the present invention is also ahead of the other three methods; from the perspective of the overall accuracy OA, the method proposed in the present invention also has the highest classification accuracy.

[0081] In summary, it can be seen that the method for integrated qualitative and quantitative detection of hydrated minerals based on Mars orbiter hyperspectral data proposed in the present invention can achieve an automated qualitative and quantitative detection process for hydrated minerals on Mars without relying on artificially constructed laboratory spectral libraries while maintaining high confidence. It has advantages in terms of the credibility of the results of hydrated mineral abundance on Mars and is applicable to Mars hyperspectral data of different resolutions.

[0082] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for detecting hydrous minerals based on Mars orbiter hyperspectral data, characterized by: include: Obtain the distribution area of ​​hydrated minerals based on Mars hyperspectral data; performing automated endmember extraction on the distribution area to obtain image endmembers; The image endmembers are inspected for spectral characteristics and unmixed based on an image endmember library to obtain a hydrous mineral abundance map.

2. The method for detecting hydrous minerals based on Mars orbiter hyperspectral detection according to claim 1, wherein: The method of obtaining the distribution area of ​​hydrated minerals based on Mars hyperspectral data refers to using a target detection method for Mars hyperspectral data, using the spectra of hydrated minerals in a reference spectral library as a guide, and obtaining high-confidence distribution areas of various types of hydrated minerals; The target detection method for Mars hyperspectral data includes: The hierarchical minimum output energy constraint method is used to detect targets in Mars hyperspectral data using the spectra of hydrous minerals in the reference spectral library as a guide.

3. The method for detecting hydrous minerals based on Mars orbiter hyperspectral detection according to claim 2, wherein: Automated end member extraction of the distribution area includes: The number of end members N is preset for the high confidence distribution areas of the various types of hydrous minerals, and the vertex component analysis method is used as the end member extraction method to obtain image end members according to the set number of end members N.

4. The method for detecting hydrous minerals based on Mars orbiter hyperspectral detection according to claim 3, wherein: The said obtaining of the water-bearing mineral abundance mapping comprises the following steps: The endmember verification method based on spectral angle and spectral absorption feature inspection removes the noise and background pixels that are incorrectly extracted from the image endmembers, and obtains high-confidence hydrous mineral endmembers in the image; An image unmixing end-element library is automatically constructed to unmix the hyperspectral data to obtain the abundance results of various hydrated minerals and the total abundance of hydrated minerals on Mars.

5. The method for detecting hydrous minerals based on Mars orbiter hyperspectral detection according to claim 4, characterized in that: Endmember inspection methods based on spectral angles include: The spectral angle matching method is used to perform global matching of the overall spectral shape. A buffer zone is established based on the absorption center position and absorption width of various types of hydrous minerals in the spectral library to constrain the local spectral absorption characteristics. Image endmembers that meet both global matching and local constraints are considered to be high-confidence hydrous mineral endmembers.

6. The method for detecting hydrous minerals based on Mars orbiter hyperspectral detection according to claim 5, characterized in that: Endmember inspection methods based on spectral absorption feature inspection include: Constructing absorption features based on the spectral library, defining the absorption range to be checked and the absorption center of the spectral library, and constructing a buffer range for the absorption range to be checked considering the spectral variability; The absorption left endpoint, absorption right endpoint and absorption center of the image endmember are located according to the gradient method. The image endmember is considered to be correctly extracted only when all three positions of the absorption interval fall within the allowable range of the reference spectrum, otherwise the corresponding image endmember is eliminated.

7. The method for detecting hydrous minerals based on Mars orbiter hyperspectral detection according to claim 6, characterized in that: The hydrous mineral abundance mapping includes using different colors to represent the range of hydrous mineral abundance on Mars.

8. A system for detecting hydrous minerals based on Mars orbiter hyperspectral data, based on the method for detecting hydrous minerals based on Mars orbiter hyperspectral data according to any one of claims 1 to 7, characterized in that: include, The region acquisition module is used to obtain the distribution area of ​​​​water-bearing minerals based on Mars hyperspectral data; Endmember extraction module, used to automatically extract endmembers from the distribution area to obtain image endmembers; The abundance mapping module is used to check the spectral characteristics of image end members and unmix them based on the image end member library to obtain the abundance mapping of water-bearing minerals.

Images