CN-120014100-B - Automatic mapping method and system for moon impact pit micro-morphology landform

CN120014100BCN 120014100 BCN120014100 BCN 120014100BCN-120014100-B

Abstract

The application provides an automatic mapping method and system for moon impact pit micro-morphology landforms, and belongs to the technical field of image data processing. The method comprises the steps of obtaining an alternative grid data set, wherein the alternative grid data set comprises lunar digital topographic data, lunar image data and topographic derivative data, the topographic derivative data comprise a plurality of topographic derivative factors, screening the alternative grid data set to determine optimal feature combinations, combining the optimal feature combinations with tags of impact pit micro-morphology subclasses to construct an impact pit micro-morphology subclass data set, inputting the impact pit micro-morphology subclass data set into a pre-constructed deep learning model to train to obtain a target model, and automatically extracting the impact pit micro-morphology subclasses in a full-month range by using the target model to finish drawing of the full-month landform. The scheme relieves the problem that the current stage of impact pit landform micro-morphology subclass mainly depends on manual extraction and has insufficient efficiency, and provides possibility for high-precision and rapid mapping of the whole-month landform.

Inventors

DENG JIAYIN
CHENG WEIMING
WANG RUIBO
LIU DANYANG
JIAO YIMENG

Assignees

中国科学院地理科学与资源研究所

Dates

Publication Date: 20260505
Application Date: 20241227

Claims (8)

1. The automatic mapping method for the micro-morphological landforms of the moon impact pit is characterized by comprising the following steps of: acquiring an alternative raster data set, wherein the alternative raster data set comprises lunar digital topographic data, lunar image data and topographic derivative data, and the topographic derivative data comprises a plurality of topographic derivative factors; screening the alternative raster data sets to determine an optimal feature combination; Combining the optimal feature combination with a tag of the impingement pit micro-morphology subclass to construct an impingement pit micro-morphology subclass dataset; inputting the impact pit micro-morphology subclass data set into a pre-constructed deep learning model for training to obtain a target model; Automatically extracting the micro-morphological subclasses of the impact pit in the whole month range by using the target model to obtain an extraction result, and further completing the drawing of the whole month landform based on the extraction result; the impact pit micro-morphology subclasses are divided into four subclass unit types, namely a central peak, a pit bottom, a pit wall and a pit edge; screening the candidate raster data set to determine an optimal feature combination, including: performing normalization pretreatment on each piece of candidate raster data in the candidate raster data set, and drawing a characteristic value distribution diagram of each piece of candidate raster data in different impact pit micro-morphology subclasses based on the normalization pretreatment result; analyzing differences of different impact pit micro-morphology subclasses on the same alternative raster data based on the characteristic value distribution diagram so as to exclude the alternative raster data with the differences smaller than a preset threshold value, and obtaining a first screening result; adopting a correlation analysis method to analyze the correlation among all terrain derivative factors in the first screening result; The method comprises the steps of carrying out secondary screening on the first screening result based on the result of correlation analysis, comprehensively considering the characteristic representation capability of different candidate characteristic data and the correlation strength among different terrain derivative data by utilizing the correlation analysis, thereby eliminating the terrain derivative data with higher overlapping degree and reserving the data with strong characteristic expression capability; and taking the characteristic combination after secondary screening as the optimal characteristic combination, wherein the optimal characteristic combination comprises moon DEM data, gradient and plane curvature.
2. The method of claim 1, wherein the terrain derivative factor comprises a macroscopic terrain factor comprising a lunar surface terrain relief; The generating step of the relief degree of the lunar surface topography is as follows: and analyzing the lunar surface relief by adopting a mean value variable point method based on the lunar digital topographic data, determining an optimal window for calculating the lunar surface relief, and calculating the topographic relief by adopting a window analysis method according to the optimal window for the lunar surface relief.
3. The method of claim 2, wherein the macro topography factor further comprises roughness, The roughness generation steps are as follows: and calculating the roughness of the lunar surface by adopting a root mean square elevation method based on the lunar digital terrain data.
4. The method of claim 1, wherein the terrain derivative factor comprises a microscopic terrain factor comprising grade, slope and curvature; the gradient, the slope direction and the curvature are all calculated based on the lunar digital terrain data.
5. The method of claim 1, wherein the impinging the tag of the pit micro-morphology subclass is generated by: acquiring first impact pit surface data, second impact pit surface data and third impact pit surface data; matching the first impact pit surface data, the second impact pit surface data and the third impact pit surface data, and drawing and finishing the impact pit micro-morphology subclasses by combining the moon digital topography data, the gradient data and the mountain shadow data to obtain labels of the impact pit micro-morphology subclasses in a vector form; The first impact pit surface data, the second impact pit surface data and the third impact pit surface data are lunar geologic maps which are respectively based on different sensor data and compiled by different teams.
6. The method of claim 5, further comprising, after obtaining the tag of the impingement pit micro-morphology subclass in vector form: And carrying out category coding and rasterization on the vector-form impact pit micro-morphology subclasses, and converting the rasterized impact pit micro-morphology subclasses into multi-channel data by adopting single-thermal coding.
7. The method of claim 1, wherein the deep learning model is constructed by modifying an input channel to be multi-channel and modifying an Upsampling module of an original U-Net to be Max Unpooling module on the basis of a deep convolutional network U-Net.
8. An automatic mapping system for micro-morphology landforms of moon impact pits, which is characterized by comprising: The system comprises a data acquisition unit, a data processing unit and a data processing unit, wherein the data acquisition unit is configured to acquire an alternative raster data set, the alternative raster data set comprises lunar digital topographic data, lunar image data and topographic derivative data, and the topographic derivative data comprises a plurality of topographic derivative factors; The data screening unit is configured to screen the alternative raster data set and determine an optimal feature combination; a training set construction unit configured to combine the optimal feature combination with the tags of the crash pit micro-morphology subclasses to construct a crash pit micro-morphology subclass dataset; the model training unit is configured to input the impact pit micro-morphology subclass data set into a pre-constructed deep learning model for training to obtain a target model; the automatic extraction unit is configured to automatically extract the micro-morphology subclasses of the impact pit in the whole month range by using the target model to obtain an extraction result, and further finish drawing of the landform of the impact pit in the whole month based on the extraction result; the impact pit micro-morphology subclasses are divided into four subclass unit types, namely a central peak, a pit bottom, a pit wall and a pit edge; screening the candidate raster data set to determine an optimal feature combination, including: performing normalization pretreatment on each piece of candidate raster data in the candidate raster data set, and drawing a characteristic value distribution diagram of each piece of candidate raster data in different impact pit micro-morphology subclasses based on the normalization pretreatment result; analyzing differences of different impact pit micro-morphology subclasses on the same alternative raster data based on the characteristic value distribution diagram so as to exclude the alternative raster data with the differences smaller than a preset threshold value, and obtaining a first screening result; adopting a correlation analysis method to analyze the correlation among all terrain derivative factors in the first screening result; The method comprises the steps of carrying out secondary screening on the first screening result based on the result of correlation analysis, comprehensively considering the characteristic representation capability of different candidate characteristic data and the correlation strength among different terrain derivative data by utilizing the correlation analysis, thereby eliminating the terrain derivative data with higher overlapping degree and reserving the data with strong characteristic expression capability; and taking the characteristic combination after secondary screening as the optimal characteristic combination, wherein the optimal characteristic combination comprises moon DEM data, gradient and plane curvature.

Description

Automatic mapping method and system for moon impact pit micro-morphology landform Technical Field The application relates to the technical field of image data processing, in particular to an automatic mapping method and system for micro-morphology landforms of moon impact pits. Background The moon surface landform type, especially the landform type of the impact pit is complex and various, and the extraction of the impact pit is an important link in the lunar landform map compiling research and is also a work needing to consume a great deal of time and cost. With the development of artificial intelligence technologies such as machine learning, neural networks and the like, the related technologies adopt supervised learning algorithms such as genetic algorithm, adaboost, support vector machine and the like, and a high-efficiency strong classifier is trained to realize automatic identification and extraction of the collision pit. However, in the prior art, the automatic extraction of the impact pit is mostly focused on extracting the boundary and position information of the impact pit, but the research of automatic extraction and segmentation of the micro-morphology subclasses of the impact pit is less, the extraction accuracy is insufficient, the precision requirement of actual drawing is difficult to meet, and the high-precision full-moon feature drawing efficiency of the micro-morphology subclasses of the impact pit is lower. Accordingly, there is a need to provide an improved solution to the above-mentioned deficiencies of the prior art. Disclosure of Invention The application aims to provide an automatic mapping method and system for micro-morphology and landforms of moon impact pits, which are used for solving or relieving the problems in the prior art. In order to achieve the above object, the present application provides the following technical solutions: The application provides an automatic mapping method for micro-morphology landforms of moon impact pits, which comprises the following steps: acquiring an alternative raster data set, wherein the alternative raster data set comprises lunar digital topographic data, lunar image data and topographic derivative data, and the topographic derivative data comprises a plurality of topographic derivative factors; screening the alternative raster data sets to determine an optimal feature combination; Combining the optimal feature combination with a tag of the impingement pit micro-morphology subclass to construct an impingement pit micro-morphology subclass dataset; inputting the impact pit micro-morphology subclass data set into a pre-constructed deep learning model for training to obtain a target model; Automatically extracting the micro-morphological subclasses of the impact pit in the whole month range by using the target model to obtain an extraction result, and further completing the drawing of the whole month landform based on the extraction result; the impact pit micro-morphology subclasses are divided into four subclass unit types of central peaks, pit bottoms, pit walls and pit edges. In one possible embodiment, the terrain derivative factor comprises a macroscopic terrain factor comprising a lunar surface terrain relief; The generating step of the relief degree of the lunar surface topography is as follows: and analyzing the lunar surface relief by adopting a mean value variable point method based on the lunar digital topographic data, determining an optimal window for calculating the lunar surface relief, and calculating the topographic relief by adopting a window analysis method according to the optimal window for the lunar surface relief. In one possible embodiment, the macro topography factor further comprises a roughness, The roughness generation steps are as follows: and calculating the roughness of the lunar surface by adopting a root mean square elevation method based on the lunar digital terrain data. In one possible embodiment, the terrain derivative factor comprises a microscopic terrain factor comprising slope, slope direction, and curvature; the gradient, the slope direction and the curvature are all calculated based on the lunar digital terrain data. In one possible implementation, the screening the candidate raster data set to determine the optimal feature combination includes: performing normalization pretreatment on each piece of candidate raster data in the candidate raster data set, and drawing a characteristic value distribution diagram of each piece of candidate raster data in different impact pit micro-morphology subclasses based on the normalization pretreatment result; analyzing differences of different impact pit micro-morphology subclasses on the same alternative raster data based on the characteristic value distribution diagram so as to exclude the alternative raster data with the differences smaller than a preset threshold value, and obtaining a first screening result; And taking the first screening result as the optimal