Automatic mapping method and system for micro-form landform of moon impact crater

Abstract

The invention provides an automatic mapping method and system for the micro-form landform of a moon impact crater, and belongs to the technical field of image data processing. Obtaining an alternative raster data set including moon digital topographic data, moon image data and topographic derivative data; the terrain derivative data comprises a plurality of terrain derivative factors; screening the alternative raster data set to determine an optimal feature combination; combining the optimal feature combination with the tag of the impact crater micro-form subclass to construct an impact crater micro-form subclass data set; inputting the impact crater micro-form subclass data set into a pre-constructed deep learning model for training to obtain a target model; and using the target model to automatically extract the micro-form subclasses of the impact craters in the whole-month range, and then completing the mapping of the whole-month landform. According to the scheme, the problems that the micro-form subclass of the landform of the impact crater at the present stage mainly depends on manual extraction and the efficiency is insufficient are solved, and high-precision and rapid mapping of the landform of the whole month is possible.

Description

Technical Field

[0001] The present application relates to the field of image data processing technology, and in particular to a method and system for automatically mapping lunar impact crater micromorphology. Background Art

[0002] The lunar surface landform types, especially the landform types of impact craters, are complex and diverse in shape. The extraction of impact craters is an important part of the research and compilation of lunar landform maps, and it is also a task that requires a lot of time and cost.

[0003] With the development of artificial intelligence technologies such as machine learning and neural networks, related technologies use supervised learning algorithms such as genetic algorithms, Adaboost, and support vector machines to train high-efficiency strong classifiers to achieve automatic identification and extraction of craters.

[0004] However, existing technologies for automatic extraction of impact craters mostly focus on extracting the boundaries and location information of impact craters, but there are fewer studies on the automatic extraction and segmentation of impact crater micromorphological subclasses. The extraction accuracy is insufficient to meet the precision requirements of actual mapping, and the efficiency of high-precision full-moon lunar mapping of impact crater micromorphological subclasses is low.

[0005] Therefore, it is necessary to provide an improved technical solution to address the above-mentioned deficiencies in the prior art. Summary of the invention

[0006] The purpose of this application is to provide a method and system for automatic mapping of lunar impact crater micromorphology to solve or alleviate the problems existing in the above-mentioned prior art.

[0007] In order to achieve the above objectives, this application provides the following technical solutions:

[0008] The present application provides a method for automatically mapping lunar impact crater micromorphology, including:

[0009] Acquire a candidate grid data set; wherein the candidate grid data set includes lunar digital terrain data, lunar image data and terrain derivative data; the terrain derivative data includes a plurality of terrain derivative factors;

[0010] Screening the candidate grid data set to determine the best feature combination;

[0011] Combining the optimal feature combination with the labels of the impact crater micromorphology subclasses to construct an impact crater micromorphology subclass dataset;

[0012] Inputting the crater micromorphology subclass dataset into a pre-built deep learning model for training to obtain a target model;

[0013] Automatically extracting the impact crater micromorphology subtypes across the entire moon using the target model to obtain extraction results; and then completing the mapping of the entire moon's landforms based on the extraction results;

[0014] The impact crater micromorphology subtypes are divided into four subtype unit types: central peak, crater bottom, crater wall and crater edge.

[0015] In a possible implementation, the terrain-derived factor includes a macro-terrain factor, and the macro-terrain factor includes the lunar surface terrain undulation;

[0016] The steps for generating the lunar surface topography are as follows:

[0017] Based on the lunar digital topographic data, the mean change point method is used to analyze the lunar surface undulation and determine the best window for calculating the lunar surface undulation; then, based on the best window for the lunar surface undulation, the window analysis method is used to calculate the terrain undulation.

[0018] In a possible implementation, the macroscopic terrain factors also include: roughness,

[0019] The roughness generation steps are as follows:

[0020] Based on the lunar digital topographic data, the roughness of the lunar surface is calculated using a root mean square elevation method.

[0021] In a possible implementation, the terrain-derived factors include micro-terrain factors, and the micro-terrain factors include slope, aspect, and curvature;

[0022] The slope, aspect and curvature are all calculated based on the lunar digital terrain data.

[0023] In a possible implementation, screening the candidate grid data set to determine the best feature combination includes:

[0024] Performing normalization preprocessing on each candidate grid data in the candidate grid data set, and drawing a characteristic value distribution map of each candidate grid data in different impact crater micromorphology subtypes based on the result of the normalization preprocessing;

[0025] Based on the characteristic value distribution diagram, the differences between different impact crater micromorphological subtypes on the same candidate grid data are analyzed to exclude the candidate grid data with differences less than a preset threshold, thereby obtaining a first screening result;

[0026] The first screening result is taken as the best feature combination.

[0027] In a possible implementation manner, after excluding candidate images with differences less than a preset threshold and obtaining a first screening result, the method further includes:

[0028] Using correlation analysis method, analyzing the correlation between each terrain-derived factor in the first screening result;

[0029] Based on the result of the correlation analysis, the first screening result is subjected to a secondary screening to obtain the optimal feature combination.

[0030] In one possible implementation, the labels of the impact crater micromorphological subclasses are generated by the following steps:

[0031] Acquire first impact crater surface data, second impact crater surface data, and third impact crater surface data;

[0032] Matching the first crater surface data, the second crater surface data, and the third crater surface data, and drawing and decorating the crater micromorphological subclasses in combination with the lunar digital terrain data, the slope data, and the mountain shadow data, so as to obtain labels of the crater micromorphological subclasses in vector form;

[0033] Among them, the first crater surface data, the second crater surface data, and the third crater surface data are lunar geological maps compiled by different teams based on different sensor data.

[0034] In a possible implementation, after obtaining the labels of the impact crater micromorphological subtypes in vector form, the method further includes:

[0035] The vector-based impact crater micromorphology subcategories are categorized and rasterized, and the rasterized impact crater micromorphology subcategories are converted into multi-channel data using one-hot encoding.

[0036] In one possible implementation, the deep learning model is constructed as follows: based on the deep convolutional network U-Net, the input channel is modified to multi-channel, and the Upsampling module of the original U-Net is modified to a MaxUnpooling module.

[0037] This embodiment also provides a lunar impact crater micromorphology automatic mapping system, including:

[0038] A data acquisition unit is configured to acquire a candidate grid data set; the candidate grid data set includes lunar digital terrain data, lunar image data and terrain derivative data;

[0039] A data screening unit, configured to screen the candidate grid data set to determine an optimal feature combination;

[0040] A training set construction unit configured to combine the optimal feature combination with the labels of the impact crater micromorphology subclasses to construct an impact crater micromorphology subclass dataset;

[0041] A model training unit is configured to input the impact crater micromorphology subclass dataset into a pre-built deep learning model for training to obtain a target model;

[0042] An automatic extraction unit is configured to automatically extract the impact crater micromorphology subtypes of the entire moon using the target model to obtain extraction results; and then complete the mapping of the impact crater landforms of the entire moon based on the extraction results;

[0043] The impact crater micromorphology subtypes are divided into four subtype unit types: central peak, crater bottom, crater wall and crater edge.

[0044] The technical solution of the embodiment of the present application has the following beneficial effects:

[0045] In the technical solution provided in this application, the best feature combination is screened from lunar digital topography data, lunar image data and topography-derived data, and combined with the labels of crater micromorphological subclasses to construct a high-quality crater micromorphological subclass dataset, and then a deep learning model is trained. The trained deep learning model (target model) is then used to automatically extract crater micromorphological subclasses to achieve high-precision rapid mapping of the entire moon's landforms. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A schematic flow chart of a method for automatically mapping lunar crater micromorphology provided according to some embodiments of the present application.

[0047] Figure 2 Schematic diagram of the lunar impact crater landform micromorphology subtypes.

[0048] Figure 3 A schematic diagram of the process of automatically extracting lunar crater micromorphological subclasses according to some embodiments of the present application.

[0049] Figure 4 A schematic diagram of terrain-derived data.

[0050] Figure 5 The distribution map of characteristic values ​​of alternative raster data in different impact crater micromorphological subtypes.

[0051] Figure 6 Schematic diagram of the results of mapping and finishing the micromorphological subtypes of impact craters.

[0052] Figure 7 A schematic diagram of the deep learning model structure provided for some embodiments of the present application.

[0053] Figure 8 A graph of the micromorphological subclass segmentation results of the Al-Biruni impact crater based on the Crater3-Net model provided for some embodiments of the present application.

[0054] Fig. 9 A graph of Jackson crater micromorphological subclass segmentation results based on the Crater3-Net model provided for some embodiments of the present application.

[0055] Description of reference numerals:

[0056] 1- crater rim, 2- crater wall, 3- crater bottom, 4- central peak, 5- lunar crust. DETAILED DESCRIPTION

[0057] The terms "first", "second", "third" and "fourth" etc. in the specification and claims of the present application and the drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0058] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0059] It should be noted that due to the lack of research on the automatic extraction and segmentation of impact crater micromorphological subcategories in the existing technology, there is currently a technical gap in the industry as to what combination of features to use in deep learning models to achieve automatic extraction of different impact crater micromorphological subcategories.

[0060] Some existing technologies also involve drawing lunar impact crater types such as ejecta, crater rims, crater walls, crater floors, and central peaks in lunar geological maps. However, the mapping process does not use deep learning technology. Instead, it converts DEM data into grayscale classification maps, matches them with high-precision lunar images, and divides the boundaries of different impact crater micromorphological subclasses by searching for the DEM maximum value or grayscale maximum value.

[0061] However, the inventors have found that the above-mentioned division process of impact crater micromorphology subclasses ignores the complexity of the lunar surface landforms, which may lead to inaccurate boundary division, thus affecting the accuracy of lunar landform mapping. Specifically, the rim of an impact crater often has irregular terrain features in multiple directions. Delineating feature points only by the maximum slope value or the highest value of the DEM may not accurately reflect the actual shape of the rim. In particular, for more complex impact craters, local highlands or low-lying areas may be produced. These terrain features may not be captured by a single maximum slope or elevation. In addition, in actual situations, the edges of different impact crater landform micromorphology subclasses are not regular linear structures, but may contain irregular shapes. Using feature point connections to determine the boundaries of each impact crater landform micromorphology subclass may not fully capture the irregular shapes, resulting in inaccurate division.

[0062] Refer to the following Figure 2 To illustrate in detail the complexity and diversity of lunar impact crater landform micromorphological subtypes.

[0063] like Figure 2 As shown in the figure, lunar impact craters are located on the lunar crust 5 and are the most common and significant geomorphic units and geotectonic markers on the lunar surface. They are numerous and varied in shape, presenting annular pit structures of varying sizes and uneven concentrations. According to their formation mechanism, the impact crater landform can be divided into four sub-unit types: central peak 4, crater bottom 3, crater wall 2, and crater rim 1. The spatial boundary and topographic features between the central peak 4 and the crater bottom 3 are significantly different. The spatial boundaries between the crater bottom 3 and the crater wall 2 and between the crater wall 2 and the crater rim 3 in the well-preserved impact crater also have obvious features. However, in the micro-morphological sub-classes of degraded and modified impact crater landforms, the topographic features and spatial boundaries between the crater bottom 3 and the crater wall 2 do not have significant separation. Given the complexity and diversity of the impact crater micro-morphological sub-classes, how to effectively extract them using deep learning models, what content the candidate feature data set should include, and what features can effectively distinguish between different micro-morphological sub-classes are one of the issues worth studying.

[0064] In addition, in some technologies, although deep learning models have been applied to the field of lunar mapping, these technologies usually only involve extracting the overall boundary and position information of the impact crater based on the lunar DEM data and image data, and do not consider how to extract the boundaries of each micromorphological subclass of the impact crater. Based on this, this embodiment attempts to select features for micromorphological subclasses of the impact crater. It not only introduces lunar DEM data and lunar image data as alternative input data for deep learning, but also adds terrain-derived data, in order to enable the deep learning model to learn the most effective features of the micromorphological subclasses of the impact crater, thereby realizing automatic recognition and extraction of high-precision micromorphological subclass boundaries.

[0065] The embodiments of the present application are described below in conjunction with the accompanying drawings.

[0066] This embodiment provides a method for automatically mapping lunar impact crater micromorphology. Figures 1 to 9 As shown, the method includes:

[0067] Step S101: Obtain a candidate raster data set.

[0068] In this embodiment, the candidate raster data set is used as a candidate feature set of the deep learning model, and the deep learning model performs feature extraction and learning through these data.

[0069] Raster data stores spatial data in the form of a grid. Each grid unit (pixel) corresponds to a specific numerical value, which represents a certain spatial attribute or observation value, also called pixel value.

[0070] A feature set refers to the input data used for model training. For deep learning, a feature set is usually a multidimensional data matrix (such as the pixel values ​​of an image, the elevation values ​​of raster data, etc.). The raster dataset is used as the input feature of the deep learning model to help the model learn the complex patterns of the micromorphology of lunar impact craters.

[0071] In this embodiment, the candidate raster data set includes lunar digital elevation model (DEM), lunar image data and terrain derivative data.

[0072] In other words, lunar DEM data, lunar image data and terrain-derived data are all stored in the form of raster data. They are used together as candidate features of deep learning models to help the model identify impact craters of different micromorphological subtypes.

[0073] Specifically, lunar digital topographic data is obtained through satellite remote sensing, probes or other space technologies. It is used to represent the digital data of the lunar surface elevation. It is an important basic data for studying the lunar landforms, geological characteristics and other related scientific research.

[0074] Topography-derived data refers to raster data derived from basic topography data such as lunar DEM data, which are used to describe impact craters of different micromorphological subtypes on the moon. These raster data can characterize the spatial distribution characteristics of the lunar landforms.

[0075] Furthermore, the terrain derivative data includes a plurality of terrain derivative factors, each of which is used to describe a certain attribute of the lunar landform. Different terrain derivative factors can describe the morphological characteristics of the lunar landform from different angles.

[0076] Lunar image data refers to image data obtained by space probes and satellite equipment, which is usually used to study the geological features, landforms and lunar environment of the lunar surface. In this embodiment, the lunar image data can specifically be optical image data. In some embodiments, the lunar image data can also be multispectral images or thermal infrared images. This embodiment is not limited to this.

[0077] In this embodiment, the lunar digital terrain data and lunar image data can be obtained through various channels, such as different space missions, satellite observations and public data platforms. Alternatively, the lunar digital terrain data can be constructed based on the laser radar, radar imaging, stereo imaging and other technologies of lunar probes and satellites to construct a three-dimensional data set reflecting the elevation of the lunar surface. The lunar image data can also be obtained through various satellite missions, probes and space telescopes. This embodiment does not limit the method of obtaining the above data.

[0078] As an example, the lunar digital terrain data and lunar image data use the latest multi-source remote sensing data of the lunar surface. The lunar image data include: data acquired by the U.S. Lunar Reconnaissance Orbiter (LRO) and data acquired by the Japanese SELenological and Engineering Explorer (SELENE); the Lunar Reconnaissance Orbiter Camera (LROC) carried by the U.S. Lunar Reconnaissance Orbiter includes a wide angle camera (WAC) and a narrow angle camera (NAC), wherein the resolution of the wide angle camera image (LROC WAC) is about 100 meters / pixel, and the resolution of the narrow angle camera image (LROC NAC) is about 0.5 meters / pixel; the image data obtained by the terrain camera (TC) carried on the Japanese kaguya (SELENE) satellite has a resolution of about 10 meters / pixel. The lunar digital terrain data include: Lunar Orbiter Laser Altimeter LOLA elevation data. LOLA is a multi-beam laser altimeter carried on the Lunar Reconnaissance Orbiter (LRO), with an emission wavelength of 1064.4nm and an emission frequency of 28Hz. LOLA provides more than 650 million lunar surface laser altimeter data for the whole month, with an elevation accuracy of about 10cm and an accuracy of about 1m. It has high-precision global coverage, and the resolution of the digital elevation model data generated by it is about 118m / pixel, and the vertical accuracy is better than 1m. The lunar digital terrain data also includes: SLDEM data, which is a higher-resolution DEM data (SLDEM2015) generated by researchers combining LOLA data and SELENETC (Terrain Camera) data. This data covers the area between 60 degrees north and south latitude of the moon, with a data resolution of about 59m / pixel and a vertical accuracy of about 3-4m.

[0079] WAC data with a resolution of 100m / pixel, LOLA digital elevation data with a resolution of 118m / pixel, SLDEM data with a resolution of 59m / pixel, hill shadow data generated based on SLDEM with a resolution of 59m / pixel, SELENETC image data and DTM data with a resolution of 10m / pixel, etc.

[0080] Step S102: Screen the candidate grid data set to determine the best feature combination.

[0081] Furthermore, in step S102, the candidate raster data set is screened to determine the best feature combination. Screening and optimizing the feature combination can help the model focus on the most important features, thereby improving the accuracy and predictive ability of the model, while accelerating the training process and reducing computing resource consumption.

[0082] Step S103, combining the best feature combination with the labels of the crater micromorphological subclasses to construct a crater micromorphological subclass dataset, thereby effectively utilizing the digital terrain data (DEM), image data, terrain-derived data of the lunar surface, and labeled crater micromorphological subclass labels to train an accurate model to identify and classify different crater micromorphological features.

[0083] Among them, the impact crater micromorphology subtypes are divided into four subtype unit types: central peak, crater bottom, crater wall and crater rim. In other words, the label contains four classification types: central peak, crater bottom, crater wall and crater rim.

[0084] Specifically, the best feature combination is combined with the label of the crater micromorphological subclass to form a training data set (i.e., the crater micromorphological subclass data set). Each sample in the training data set corresponds to an area covered by an crater micromorphological subclass in spatial position, and each sample consists of a feature vector composed of the best feature combination and its corresponding label of the crater micromorphological subclass.

[0085] Step S104: input the crater micromorphology subclass dataset into a pre-built deep learning model for training to obtain a target model.

[0086] The purpose of step S104 is to use a deep learning model to train a classification model for crater micromorphological subclasses, based on the feature vectors and labels in the crater micromorphological subclass dataset, so that the model can automatically learn and distinguish different crater micromorphological features, and then accurately classify and identify the lunar data to be predicted.

[0087] During training, a certain proportion of samples in the crater micromorphology subclass dataset can be randomly selected as the training set, and then divided into a validation set and a test set.

[0088] Furthermore, in order to prevent gradient diffusion and gradient vanishing during the model training process, each type of input data in the crater micromorphology subclass dataset is normalized before model training.

[0089] The deep learning model may be a fully convolutional neural network (FCN), U-Net, UNet++, SegNet based on the encoder-decoder architecture, a residual network (ResNet), and Google's DeepLab series of models. This embodiment does not limit the specific selection of the deep learning model.

[0090] Step S105: Use the target model to automatically extract the impact crater micromorphological subtypes across the entire moon to obtain extraction results; and then complete the mapping of the entire moon's landforms based on the extraction results.

[0091] The trained target model is used to automatically process data from the entire moon, quickly and efficiently extract crater micromorphological subtypes, and construct a detailed lunar surface topography map based on the extraction results, showing craters of different micromorphological subtypes and their characteristic areas, greatly improving the efficiency and accuracy of crater micromorphological subtype identification and topography mapping.

[0092] Preferably, the terrain-derived factors include macro-terrain factors, and the macro-terrain factors include the lunar surface terrain undulation.

[0093] In this embodiment, the terrain relief is also called topographic relief and relative height.

[0094] The steps for generating lunar surface topography are as follows:

[0095] Based on the lunar digital topographic data, the mean change point method is used to analyze the lunar surface undulation and determine the optimal window for calculating the lunar surface undulation. Then, based on the optimal window for calculating the lunar surface undulation, the window analysis method is used to calculate the terrain undulation.

[0096] The terrain relief is the elevation difference within a given regional unit (analysis window). The relief will change within a certain range as the analysis window changes, and it has a certain scale dependence.

[0097] Among them, the analysis windows mainly include rectangular windows, circular windows, ring windows, fan-shaped windows, etc.

[0098] For example, the optimal window may be a circular window. Using a circular window instead of a rectangular window to determine the optimal window for the lunar surface roughness can better fit the shape of the impact crater, which is conducive to improving the accuracy of the calculation of the roughness.

[0099] In this embodiment, the mean change point method is used to analyze the lunar surface fluctuation to determine the size of the optimal window for calculating the lunar surface fluctuation.

[0100] Since the lunar surface is undulating and the amount of data calculated for the entire month is large, in order to accurately determine the optimal window for undulation, this embodiment further uses the mean change point method twice for calculation. The first time the mean change point method is used to estimate the possible window for the optimal undulation, for example, the starting window is a circular window with a radius of 5 pixels, and the step size is 5. The average undulation within 500 pixels and its corresponding statistics are calculated to obtain the number of pixels corresponding to the possible window for the optimal undulation; the second time the mean change point method is used to perform an accurate analysis of the optimal window, for example, the starting window is 1 pixel, the step size is 1, and the calculation is twice the number of pixels corresponding to the possible window for the optimal undulation obtained by the preliminary analysis to obtain the exact value of the number of pixels in the optimal window.

[0101] Preferably, the macro-topographic factors also include: roughness, and the roughness generation steps are as follows: based on the lunar digital topographic data, the roughness of the lunar surface is calculated using the root mean square elevation method.

[0102] Specifically, roughness is used to describe how the elevation of the lunar surface changes with the horizontal scale. In other words, the roughness of the lunar surface quantitatively describes the topographic undulations at the horizontal scale, and this factor can, to a certain extent, reveal the spatial differentiation characteristics of the lunar surface landforms.

[0103] The root mean square (RMS) method is used to calculate the roughness of the lunar surface. By calculating the deviation and square average of the elevation values, the influence of accidental noise or local irregularities can be reduced. Even if there are some outliers or small noise in the data, the RMS value can remain relatively stable. At the same time, the RMS value directly reflects the roughness of the area. It is an intuitive and simple calculation method suitable for the calculation of large-scale raster data on the lunar surface.

[0104] Preferably, the terrain-derived factors include micro-terrain factors, which include slope, aspect and curvature; the slope, aspect and curvature are all calculated based on lunar digital terrain data.

[0105] Among them, the slope of any point on the lunar surface refers to the angle between the tangent plane passing through the point and the horizontal ground, which indicates the degree of inclination of the local surface slope. The size of the slope directly affects the scale and intensity of surface material flow and energy conversion.

[0106] Furthermore, the maximum average method is used to calculate the slope data for the whole month based on the lunar digital topographic data.

[0107] Curvature includes profile curvature and plane curvature. Profile curvature is the rate of change of slope, and plane curvature is the rate of change of slope direction. Curvature can reveal the curved features of the lunar surface and help understand the changing process of the lunar surface morphology.

[0108] In this embodiment, the slope aspect and curvature are calculated based on the lunar digital terrain data.

[0109] Figure 4 A schematic diagram of terrain-derived data is shown, where (a) is relief amplitude (RA), (b) is slope, (c) is profile curvature, (d) is aspect, (e) is plan curvature, and (f) is root mean square height. It can be seen from the figure that different terrain-derived factors can describe the micromorphological subclasses of impact craters from different aspects and can be used as alternative input data for deep learning models.

[0110] Preferably, the candidate grid data set is screened to determine the best feature combination, including:

[0111] Performing normalization preprocessing on each candidate grid data in the candidate grid data set, and drawing a characteristic value distribution map of each candidate grid data in different impact crater micromorphology subtypes based on the result of the normalization preprocessing;

[0112] Based on the eigenvalue distribution map, the differences between different crater micromorphological subtypes on the same candidate grid data are analyzed to exclude the candidate grid data with differences less than a preset threshold, and the first screening result is obtained; the first screening result is used as the optimal feature combination.

[0113] In the above steps, we first select multiple typical impact craters of different ages and morphological characteristics, and then normalize and preprocess each candidate raster data in the candidate raster data set. Then, we draw the distribution characteristics of each candidate raster data in different impact crater micromorphological subtypes to obtain the eigenvalue distribution map. Based on the eigenvalue distribution map, we exclude features with little difference to improve the recognition effect of the model, reduce computational overhead, and improve the interpretability of the model.

[0114] For example, Figure 5 As shown in the eigenvalue distribution diagram, LDEM represents the lunar DEM data, Aspect is the slope direction, Plan_Cur is the plane curvature, Profile_Cur is the profile curvature, Slope is the slope, RA is the terrain undulation, MAS is the roughness represented by the root mean square height, Shade is the mountain shadow, and WAC is the lunar image data. Figure 5It can be seen that the distribution differences of the different micromorphological subclasses of impact craters on different candidate raster data are not the same. Among them, in the four candidate raster data of plane curvature, slope, undulation, and terrain roughness, the four micromorphological subclasses of impact craters have better discrimination, followed by the discrimination in lunar DEM data, mountain shadow, lunar image data, and profile curvature data. In the aspect data, there is almost no difference in the characteristics of the micromorphological subclasses of impact craters. Based on this, the aspect can be excluded from the candidate raster data set to obtain the first screening result.

[0115] To further improve the sensitivity of the feature, in some embodiments, after excluding candidate images with differences less than a preset threshold and obtaining the first screening result, the method further includes: using a correlation analysis method to analyze the correlation between the terrain-derived factors in the first screening result;

[0116] Based on the results of the correlation analysis, the first screening results were screened again to obtain the best feature combination.

[0117] The results of correlation analysis between different terrain-derived data in the impact crater micromorphology subtypes show that although elevation is highly correlated with most terrain factors, considering the spatial structure expressed by elevation information, elevation data is retained as input channel data; in addition, slope is highly correlated with terrain relief, mountain shadow, profile curvature, etc., so slope is selected as input channel data, and terrain relief, mountain shadow and profile curvature with high correlation are excluded. Plane curvature has a low correlation with slope, terrain relief, lunar surface roughness, profile curvature, etc., and plane curvature is considered as an alternative input data.

[0118] By using correlation analysis, we comprehensively consider the feature representation capabilities of different candidate feature data and the strength of the correlation between different terrain-derived data, thereby eliminating terrain-derived data with high overlap, retaining data with strong feature expression capabilities, and reducing the amount of calculation. In addition, we use correlation analysis as a supplement to the eigenvalue distribution map analysis, and screen the candidate feature set through two screening methods, which can ensure that the feature vector of the input model can establish an effective mapping relationship with the model output result, ensuring that the deep learning model can correctly identify the different micromorphological sub-class features of the impact crater, and thus obtain accurate recognition results.

[0119] In an optional embodiment, the best feature combination can be: lunar DEM data, slope and plane curvature. The above three features will be used as input channel data of the deep learning model to participate in the subsequent model training and automatic extraction of impact crater micromorphological subclasses.

[0120] Considering that the deep learning algorithm requires enough samples, and summarizes, learns, and finds different features of impact crater micromorphological subclasses to train the network, and segment the impact crater micromorphological subclasses based on the trained model, it is necessary to obtain label data in the training sample set. Preferably, the labels of the impact crater micromorphological subclasses are generated by the following steps:

[0121] Acquire first impact crater surface data, second impact crater surface data, and third impact crater surface data;

[0122] Match the first crater surface data, the second crater surface data, and the third crater surface data, and draw and decorate the crater micromorphological subclasses in combination with the lunar digital terrain data, slope data, and mountain shadow data to obtain the crater micromorphological subclass labels in vector form;

[0123] Among them, the first crater surface data, the second crater surface data, and the third crater surface data are lunar geological maps compiled by different teams based on different sensor data.

[0124] In this embodiment, the actual boundaries of the impact crater micromorphological subtypes are drawn mainly based on manual visual interpretation of expert knowledge, and the base map is derived from different sensor data and lunar geological maps compiled by different teams.

[0125] Due to the differences between the data obtained by different sensors and the images they generate, there are certain positional deviations in the crater boundaries drawn based on different data. In addition, the crater details expressed in the lunar geological maps compiled by different teams are not the same. For example, the crater surface data provided by some lunar geological maps only draw the actual boundaries of the craters, that is, the lines connecting the points where the slope changes at the crater mouth, and do not distinguish between the micromorphological subtypes of the craters. In the crater surface data provided by other lunar geological maps, for craters with larger diameters, the crater bottom subtypes are identified, and no distinction is made between the crater wall and the crater edge. In some data, there is an overlapping relationship between the craters of each age and no cutting is done. Therefore, it is necessary to match the crater surface data from different sources, and to combine the lunar digital terrain data, slope data, and mountain shadow data to perform fine drawing and finishing of the crater micromorphological subtypes in order to improve the drawing accuracy of the label data and thus improve the quality of the training data set.

[0126] Figure 6 The following figure shows the results of mapping and finishing the impact crater micromorphology subtypes. (a) shows the impact crater micromorphology subtype area mapped in a lunar geological map, and (b), (c), and (d) show the mapping and finishing results of the impact crater micromorphology subtypes under different base maps.

[0127] In addition, it should be noted that for impact craters that are very old and have degraded due to space weathering, their micromorphological subtypes cannot be identified and classified, so they are excluded when making samples to further improve the quality of training samples.

[0128] Preferably, after obtaining the labels of the impact crater micromorphological subtypes in vector form, the method further includes:

[0129] The vector-based impact crater micromorphology subcategories are categorized and rasterized, and one-hot encoding is used to convert the rasterized impact crater micromorphology subcategories into multi-channel data.

[0130] For the label data Y of the impact crater micromorphology subclass, it corresponds to the category value. In this embodiment, there are four target categories: central peak, crater bottom, crater wall, and crater edge. Together with background pixels, there are five categories in total. In order to input the label data into the deep learning model, category encoding, rasterization and one-hot encoding are required to process the label data in vector form.

[0131] Exemplarily, category encoding includes: converting the central peak, pit bottom, pit wall, and pit edge into numerical types, encoding them as 1, 2, 3, and 4, and encoding the remaining pixels as 0; rasterization includes: rasterizing the vector data at the same resolution scale as the corresponding raster data to obtain the corresponding label grid, and the raster value range is 0 to 4; one-hot encoding includes: semantic segmentation is essentially a pixel-level classification task, and the research in this embodiment corresponds to a multi-classification task, so the data after rasterization in the previous step is one-hot encoded, converting it from single-channel data to 5-channel data, and the corresponding data in each channel has only values ​​0 and 1, where 1 represents a pixel of this category and 0 represents a pixel of a non-category.

[0132] Preferably, the deep learning model is constructed as follows: based on the deep convolutional network U-Net, the input channel is modified to multi-channel, and the Upsampling module of the original U-Net is modified to a Max Unpooling module.

[0133] The traditional U-Net structure has only one input channel. However, in the study of lunar landforms, the lunar landform features are a complex combination. Data based on one channel cannot accurately express the characteristics of the landform type. For example, image data can reflect the texture characteristics of the landform, but cannot represent the terrain characteristics of the elevation data. At the same time, the automatic classification of lunar landforms requires the input of different data sets to express the various characteristics of the landform type. Therefore, in this embodiment, it is necessary to improve the traditional U-Net structure and construct a U-Net network with multiple input channels. For the convenience of description, the improved U-Net network is called Crater3-Net.

[0134] like Figure 7 As shown in the figure, compared with the traditional U-Net network, the improvements of Crater3-Net are:

[0135] (1) The input channels are modified to be multi-channel (e.g., 3 channels) so that the network can fully utilize the characteristics of the lunar DEM and its derivative data;

[0136] (2) The original Upsampling module in the traditional U-Net network is modified into a Max Unpooling module to adapt to the segmentation characteristics of the lunar impact crater micromorphology and keep the original main feature positions unchanged.

[0137] For example, the execution flow of the Crater3-Net network is as follows:

[0138] (1) The input data (i.e., the best feature combination) is converted into a multi-channel tensor after image processing and enhancement, for example, a tensor with 3 channels and a height and width of 512.

[0139] (2) Perform convolution and pooling operations on the multi-channel tensor generated above, where the pooling uses the maximum pooling method and the step size is set to 2, that is, each time after the pooling operation, the size of the feature map is reduced by half. At the same time, in order to maintain the same amount of information, the number of channels is doubled after the convolution operation. Add BatchNorm and PReLU activation function layers after each convolution layer.

[0140] (3) Repeat the above step (2) until the size of the feature map becomes 32*32, and the number of channels is 256.

[0141] It should be noted that during the maximum pooling operation in steps (2) and (3), the position of the maximum value should be recorded.

[0142] (4) Perform the maximum unpooling operation on the result obtained above, gradually restore the size of the feature map of the left half of the network, and perform channel splicing on the feature maps of corresponding sizes.

[0143] (5) The final output result of 5 channels is obtained through convolution, and its size is 512*512. During training, this output result and label data are used to calculate the loss, and then the gradient is back-propagated to update the network parameters; during inference, the maximum value index operation is performed in the channel dimension to obtain the final segmentation result.

[0144] Considering the complexity and diversity of the morphological features of impact craters, some impact craters may not have a central peak, or some impact craters may not have a bottom, and there may be a sample imbalance problem after one-hot encoding (the proportion of target pixels is smaller than that of background pixels), so DiceLoss is selected as the loss function when training the deep learning model. The specific calculation formula of DiceLoss can be executed with reference to the existing technology, and will not be repeated here.

[0145] In order to better train the model and enhance the range of the model, the following methods are used for image processing and enhancement during the model training process:

[0146] (1) Scaling: To facilitate network training, U-Net networks generally require a fixed image size during training. Considering the data of the impact craters selected in this study, the grid data within the corresponding range of each impact crater is interpolated to a size of 512*512. In order to avoid introducing new noise, the interpolation process uses the nearest neighbor interpolation method.

[0147] (2) Horizontal flipping: During training, the input data and the corresponding label data are horizontally flipped with a probability of 0.5.

[0148] (3) Vertical flipping: During training, the input data and the corresponding label data are horizontally flipped with a probability of 0.5. In particular, when the model is used for inference, only the image size is scaled, and the size and coordinate information of the original data is recorded before the data is input into the model. The output data of the model is then restored using this information to ensure that it has the same size and geographic location as the original data, so as to facilitate later vectorization and mapping.

[0149] Furthermore, the training process of the Crater3-Net network adopts a strategy of dynamically adjusting the learning rate. Specifically, the initial learning rate of the model is set to 0.001, and the Reduce LR On Plateau learning rate adjustment strategy is adopted. This strategy dynamically adjusts the learning rate of the model according to the loss change of the validation set when training the model. When the loss on the validation set does not change for N consecutive epochs, the learning rate is reduced by S times. When training the model, N is set to 3 and S is set to 0.5.

[0150] To verify the performance of the Crater3-Net network provided in this embodiment, the following four models are trained using the same model parameter settings, and the results are evaluated and compared:

[0151] (1) Crater3-Net: The model is obtained based on the above network architecture, that is, the input channels include three data channels: elevation data, plane curvature and slope, and the upsampling module adopts Max Unpooling.

[0152] (2) Crater1-Net: This model is obtained based on the above network architecture, but only the elevation data is retained in the input channel, and the upsampling module uses Max Unpooling.

[0153] (3) U3-Net: The original U-Net architecture only modifies the input channels to 3, including three data channels: elevation data, plane curvature, and slope data. The upsampling uses the nearest neighbor interpolation method.

[0154] (4) U1-Net: The original U-Net architecture, the input channels include elevation data, plane curvature and slope data, and the upsampling uses the nearest neighbor interpolation method.

[0155] The mean Intersection over Union (mIoU) and mean pixel accuracy (MPA) of the above four models are calculated as evaluation indicators, and the learning rate change curves of the four models, the change curves of Loss on different training sets and different validation sets, and the change curves of each evaluation indicator are plotted during the training process.

[0156] The results show that the time of the first learning rate decay of different models is not consistent. The Crater1-Net model decays the fastest, followed by U1-Net, and the Crater3-Net model decays the slowest. Under the same random seed, the timing of learning rate decay of different models is different, indicating that the learning rates of different models are different. After the four models start training, as the number of traversal data increases, the loss function value decreases rapidly. After about 50 epochs of training, the four models have basically reached a convergence state. From the evaluation indicators mIOU and MPA, the final effects of the four models are different: after convergence, the mIOU of the four models on the validation set is close to 0.8. Among them, the Crater3 model has the best effect, followed by U3-Net, then U1-Net, and finally Crater1-Net; and the MPA indicator also shows similar results. After the convergence of each model, its MPA value is around 0.95. Among them, the Crater3-Net model has the best effect, followed by U3-Net, then U1-Net and Crater1-Net. From the change of mIOU MPA index on the validation set, when using elevation data, plane curvature and slope data as input data at the same time, the effect is better than using only elevation data as input data; from another perspective, when the input data is 3 channels, in the decoder part, the effect of using MaxUnpooling for upsampling is better than using the nearest neighbor interpolation method, and when the input data is single channel, the effect of using the nearest neighbor interpolation method for upsampling is slightly better than using Max Unpooling for upsampling. But in general, whether from the perspective of mIOU or MPA, the Crater3-Net model has the best segmentation effect.

[0157] Furthermore, the above four trained models are used to segment the micromorphological subclasses of impact craters on the test dataset, and their evaluation indicators are statistically analyzed to obtain the following table:

[0158] Table 1 Segmentation accuracy comparison

[0159]

[0160] It can be seen from the above table that the method provided in this embodiment has the highest segmentation accuracy on the test data set, which is close to 0.8.

[0161] In addition, in order to further verify the practical effect of this embodiment, impact craters of different morphological types (such as the Cauchy crater and the Al-Biruni crater) were selected to compare the segmentation effects, and the segmentation results were evaluated from qualitative and quantitative perspectives, as shown in Table 2.

[0162] Table 2 Comparison of segmentation accuracy of crater micromorphology subtypes of different morphological types

[0163]

[0164] Figure 8 The results of the micromorphological subclass segmentation of the Al-Biruni impact crater are shown. Figure 8 The results show that: whether it is the Cauchy crater or the Al-Biruni crater, the overall segmentation effect of the micromorphological subclasses obtained based on the Crater3-Net model is good. The segmentation accuracy of the bottom and wall of the Cauchy crater is high, with an MPA of up to 0.9, while the segmentation accuracy of the rim is relatively low, about 0.84. The segmentation results of the Al-Biruni crater include the bottom, wall and rim, with cPAs of 0.99, 0.91 and 0.96 respectively, and the MPA is close to 0.96.

[0165] The transitional crater Hoffmeister N is located in the highland area on the far side of the moon and was formed in the Neptunian period. In its northern region, there is an Neptunian crater Hoffmeister, and there is a partial overlap between the two. When segmenting based on the Crater3-Net model, the micromorphological subtypes of the craters Hoffmeister N and Hoffmeister can be well segmented, and the overlap between the two is cut. The segmentation accuracy of the crater bottom, crater wall and crater edge is high, but there are also some mis-segmentation problems. For example, there is an impact crater superimposed on the impact crater Hoffmeister, and the model segments it as a central peak.

[0166] Jackson crater is a typical complex crater in the highland area. Fig. 9 The Jackson crater micromorphological subclass segmentation result diagram is shown. It can be clearly seen from the figure that the segmentation effect of the crater wall and the crater edge in the micromorphological subclass is better than that of the crater bottom and the central peak. Combined with Table 2, it can be seen that for the segmentation of complex craters, the segmentation accuracy of the crater wall and the crater edge is relatively high, and its cPA is around 0.9, while the segmentation accuracy of the crater wall and the central peak is relatively low, and its cPA is 0.70 and 0.63 respectively. In addition, this paper converts the segmentation results into vector data and compares them with the manually drawn sample data. According to the WAC and slope data, the vector data of the Jackson crater segmentation results more accurately fits the boundaries of the crater bottom and the central peak. From another perspective, when the errors of manually labeled samples are large, the model Crater3-Net constructed in this embodiment can correct the errors of manual drawing to a certain extent to obtain better boundary results for the micromorphological subclasses of the crater.

[0167] The trained Crater3-Net model (target model) is used to automatically extract the micromorphological subtypes of impact craters across the entire moon and obtain the extraction results.

[0168] In this embodiment, the Crater3-Net model is used to obtain all the lunar craters with a direct area larger than 5 km and segment them. Based on the segmentation results, a database of crater micromorphology subtypes is obtained to complete the mapping of the lunar crater landforms.

[0169] Refer to the following Figure 3 The process of automatic extraction of lunar crater micromorphological subclasses is further described in an exemplary manner.

[0170] Figure 3 Schematic diagram of the process of automatically extracting lunar impact crater micromorphology subtypes according to some embodiments of the present application. Figure 3 As shown in the figure, the automatic extraction of lunar impact crater micromorphological subtypes can be performed as follows:

[0171] Dataset construction: including the selection of input data, extraction of terrain factors, selection of raster data sets, image screening to determine the best feature combination, sample annotation, and construction of crater micromorphology subclass dataset.

[0172] Classification data preparation: Divide the impact crater micromorphology subclass dataset into training data, test data, and validation data.

[0173] Data preprocessing: including data resampling, image cutting and image enhancement. Here, the image refers to the raster data in the crater micromorphology sub-category data.

[0174] Model construction: Build a multi-channel Crater3-Net network based on the U-Net network.

[0175] Training process: Input the training data into the Crater3-Net semantic segmentation network, calculate the loss function, and adjust the learning rate until the model converges to obtain the optimal parameters of the Crater3-Net model.

[0176] Model evaluation: Based on the test data and verification data, the converged Crater3-Net model is compared and evaluated for accuracy with the Crater1-Net model, U3-Net model, and U1-Net model. The evaluation indicators include mIOU, MPA, etc.

[0177] After the above steps, the automatic extraction and segmentation of lunar impact crater micro-morphological sub-types are realized. On this basis, the mapping of the entire lunar landform is completed.

[0178] Taking the impact landform mapping of the Chang'e-5 landing area as an example, there are 197 types of impact crater landforms classified into micromorphological subcategories, including the bottom of impact craters with small undulations at low altitudes in the lunar seas, the walls of impact craters with large undulations at extremely high altitudes in the highlands, and the rim of impact craters with extremely low altitudes and small undulations in the Aiken impact crater in the Antarctic.

[0179] In summary, impact craters are typical lunar landforms. Due to the diversity and complexity of impact crater morphology, boundary extraction requires expert knowledge assistance. At present, the automatic extraction and identification of impact craters often only determine the location information and diameter of the impact craters, and there are few studies on the actual boundary extraction of impact crater micromorphological subclasses. In order to make up for the above-mentioned technical deficiencies, in this embodiment, in the training data set construction stage, a variety of terrain-derived factors are screened as candidate features of the model to determine the optimal feature combination of the model. The model provides high-quality training data to ensure that the model can learn the characteristics of the impact crater micromorphological subclasses. Based on U-Net, a deep learning network segmentation model Crater3-Net for impact crater landform micromorphological subclasses is constructed, and impact craters with a diameter greater than 5 km in the entire moon are automatically segmented and extracted to form a database of impact crater micromorphological subclasses with a diameter greater than 5 km in the entire moon, which alleviates the problem that the current impact crater landform micromorphological subclasses mainly rely on manual extraction and are inefficient, and provides the possibility for high-precision rapid mapping of the entire moon of lunar impact landforms, thereby laying the foundation for further research on lunar landforms.

[0180] Based on the same inventive concept, this embodiment also provides a lunar impact crater micromorphology automatic mapping system, the system comprising:

[0181] A data acquisition unit is configured to acquire a candidate raster data set; the candidate raster data set includes lunar digital terrain data, lunar image data and terrain derivative data;

[0182] A data screening unit configured to screen the candidate grid data set to determine the best feature combination;

[0183] A training set construction unit configured to combine the best feature combination with the labels of the impact crater micromorphology subclasses to construct an impact crater micromorphology subclass dataset;

[0184] A model training unit is configured to input the impact crater micromorphology subclass dataset into a pre-built deep learning model for training to obtain a target model;

[0185] An automatic extraction unit is configured to automatically extract the impact crater micromorphology subtypes of the entire moon using the target model to obtain extraction results; and then complete the mapping of the impact crater landforms of the entire moon based on the extraction results;

[0186] Impact crater micromorphology subtypes are divided into four subtype unit types: central peak, crater bottom, crater wall and crater rim.

[0187] The automatic mapping system for lunar impact crater micromorphology provided in this embodiment can implement the steps and processes of the automatic mapping method for lunar impact crater micromorphology provided in any of the above embodiments, and achieve the same technical effects, which will not be repeated here one by one.

[0188] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for automatically mapping lunar impact crater micromorphology, characterized in that: include: Acquire a candidate grid data set; wherein the candidate grid data set includes lunar digital terrain data, lunar image data and terrain derivative data; the terrain derivative data includes a plurality of terrain derivative factors; Screening the candidate grid data set to determine the best feature combination; Combining the optimal feature combination with the labels of the impact crater micromorphology subclasses to construct an impact crater micromorphology subclass dataset; Inputting the crater micromorphology subclass dataset into a pre-built deep learning model for training to obtain a target model; Automatically extracting the impact crater micromorphology subtypes across the entire moon using the target model to obtain extraction results; and then completing the mapping of the entire moon's landforms based on the extraction results; The impact crater micromorphology subtypes are divided into four subtype unit types: central peak, crater bottom, crater wall and crater edge.

2. The method according to claim 1, characterized in that The terrain-derived factors include macro-terrain factors, and the macro-terrain factors include the lunar surface terrain undulation; The steps for generating the lunar surface topography are as follows: Based on the lunar digital topographic data, the mean change point method is used to analyze the lunar surface undulation and determine the best window for calculating the lunar surface undulation; then, based on the best window for the lunar surface undulation, the window analysis method is used to calculate the terrain undulation.

3. The method according to claim 2, characterized in that The macroscopic terrain factors also include: roughness, The roughness generation steps are as follows: Based on the lunar digital topographic data, the roughness of the lunar surface is calculated using a root mean square elevation method.

4. The method according to claim 1, characterized in that: The terrain-derived factors include micro-terrain factors, and the micro-terrain factors include slope, aspect and curvature; The slope, aspect and curvature are all calculated based on the lunar digital terrain data.

5. The method according to claim 1, characterized in that Screening the candidate grid data set to determine the best feature combination includes: Performing normalization preprocessing on each candidate grid data in the candidate grid data set, and drawing a characteristic value distribution map of each candidate grid data in different impact crater micromorphology subtypes based on the result of the normalization preprocessing; Based on the eigenvalue distribution diagram, the differences between different impact crater micromorphology subtypes on the same candidate grid data are analyzed to exclude the candidate grid data with differences less than a preset threshold, thereby obtaining a first screening result.

6. The method according to claim 5, characterized in that After excluding candidate images whose differences are less than a preset threshold and obtaining the first screening result, the following steps are also included: Using correlation analysis method, analyzing the correlation between each terrain-derived factor in the first screening result; Based on the results of the correlation analysis, a secondary screening is performed on the first screening results; The first screening result is used as the optimal feature combination or the feature combination after the second screening is used as the optimal feature combination.

7. The method according to claim 1, characterized in that The labels of the crater micromorphological subclasses are generated by the following steps: Acquire first impact crater surface data, second impact crater surface data, and third impact crater surface data; Matching the first crater surface data, the second crater surface data, and the third crater surface data, and drawing and decorating the crater micromorphological subclasses in combination with the lunar digital terrain data, the slope data, and the mountain shadow data, so as to obtain labels of the crater micromorphological subclasses in vector form; Among them, the first crater surface data, the second crater surface data, and the third crater surface data are lunar geological maps compiled by different teams based on different sensor data.

8. The method according to claim 7, characterized in that After obtaining the labels of the crater micromorphological subclasses in vector form, also include: The vector-based impact crater micromorphology subcategories are categorized and rasterized, and the rasterized impact crater micromorphology subcategories are converted into multi-channel data using one-hot encoding.

9. The method according to claim 1, characterized in that: The deep learning model is constructed as follows: based on the deep convolutional network U-Net, the input channel is modified to multi-channel, and the Upsampling module of the original U-Net is modified to a Max Unpooling module.

10. An automatic mapping system for lunar impact crater micromorphology, characterized in that: include: A data acquisition unit is configured to acquire a candidate grid data set; the candidate grid data set includes lunar digital terrain data, lunar image data and terrain derivative data; A data screening unit, configured to screen the candidate grid data set to determine an optimal feature combination; A training set construction unit configured to combine the optimal feature combination with the labels of the impact crater micromorphology subclasses to construct an impact crater micromorphology subclass dataset; A model training unit is configured to input the impact crater micromorphology subclass dataset into a pre-built deep learning model for training to obtain a target model; An automatic extraction unit is configured to automatically extract the impact crater micromorphology subtypes of the entire moon using the target model to obtain extraction results; and then complete the mapping of the impact crater landforms of the entire moon based on the extraction results; The impact crater micromorphology subtypes are divided into four subtype unit types: central peak, crater bottom, crater wall and crater edge.

Images