CN115272769A - Automatic moon impact pit extraction method and device based on machine learning - Google Patents
Automatic moon impact pit extraction method and device based on machine learning Download PDFInfo
- Publication number
- CN115272769A CN115272769A CN202210956296.9A CN202210956296A CN115272769A CN 115272769 A CN115272769 A CN 115272769A CN 202210956296 A CN202210956296 A CN 202210956296A CN 115272769 A CN115272769 A CN 115272769A
- Authority
- CN
- China
- Prior art keywords
- impact
- candidate
- pit
- area
- identified
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 title claims abstract description 49
- 238000000605 extraction Methods 0.000 title claims abstract description 43
- 238000010801 machine learning Methods 0.000 title claims abstract description 26
- 238000000034 method Methods 0.000 claims abstract description 46
- 238000004422 calculation algorithm Methods 0.000 claims description 16
- 238000011160 research Methods 0.000 abstract description 3
- 238000012549 training Methods 0.000 description 25
- 238000004458 analytical method Methods 0.000 description 24
- 238000007637 random forest analysis Methods 0.000 description 15
- 238000010586 diagram Methods 0.000 description 9
- 230000008569 process Effects 0.000 description 9
- 238000012360 testing method Methods 0.000 description 7
- 238000009826 distribution Methods 0.000 description 5
- 230000015556 catabolic process Effects 0.000 description 3
- 238000010276 construction Methods 0.000 description 3
- 238000006731 degradation reaction Methods 0.000 description 3
- 230000000877 morphologic effect Effects 0.000 description 3
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012876 topography Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
- G06V10/763—Non-hierarchical techniques, e.g. based on statistics of modelling distributions
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Probability & Statistics with Applications (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a moon impact pit automatic extraction method and device based on machine learning, and belongs to the technical field of geographical research. The automatic moon impact pit extraction method comprises the following steps: acquiring DEM data of an area to be identified on the moon; identifying the area to be identified according to a pre-constructed grid-level classifier and DEM data of the area to be identified to obtain candidate impact pits; obtaining a radial elevation profile of the candidate impact pit according to the candidate impact pit and DEM data of the area to be identified; and identifying the candidate meteorite crater according to a pre-constructed object-level classifier and a radial elevation profile of the candidate bumping crater to obtain the bumping crater. The device comprises: the device comprises an acquisition module, a first obtaining module, a second obtaining module and a third obtaining module. According to the technical scheme, the machine learning method is combined with the geomorphic elements, and a grid-oriented and object-oriented combined method is used in the machine learning, so that the identification accuracy of the impact pits is improved.
Description
Technical Field
The invention belongs to the technical field of geographical research, and particularly relates to a moon impact pit automatic extraction method and device based on machine learning.
Background
Impact pits, the most prominent topographical feature of the lunar surface, represent a relatively complete record of the history of impact since its formation. In the whole history process of the moon, the characteristics of the size, the form, the distribution and the like of the impact pits are important for researching the evolution history of the moon. Therefore, impact pit identification is the most fundamental work in impact pit research, and results are obtained from manual interpretation to automatic extraction.
The extraction method of the impact pit can be divided into two categories of manual visual identification and automatic extraction. The former method for identifying the impact pits is simple, the result is relatively accurate, but a large number of impact pits need to be identified when the problems of impact pit density and the like are researched, and the manual extraction method is time-consuming and labor-consuming. The principles of automatic identification mainly fall into the following four categories: the first category is feature matching based methods. In some of these methods, it is proposed to automatically identify the impact pit by using parameters such as morphological characteristics of a lunar surface annular structure. The second category is methods based on image transformation and segmentation. In the method, some proposals adopt an object-oriented method to extract the moon surface impact pits, and select characteristic values of 'adjacent phase contrast' and 'length-width ratio' to identify the impact pits. The third category is methods based on geographic information fusion. In the methods, some methods have proposed 3 automatic extraction methods of hole filling, object-oriented classification and hole filling-object-oriented classification, and bump-pit extraction tests are performed on a Digital Elevation Model (DEM), which shows that the hole filling object-oriented method has higher extraction precision. The fourth category is machine learning based methods. In the methods, some proposals establish a trainable algorithm to extract and identify impact pits with different sizes by using the theory of machine learning and computer vision for reference. Compared with the first three types of automatic identification methods of the impact pits, the fourth type of machine learning method abandons the limitation based on two-dimensional image data, trains a machine learning classifier by using the traditional impact pit map and the spatial structure information knowledge obtained from digital terrain analysis, overcomes the defects based on terrain analysis and greatly improves the identification efficiency.
Impact craters are not simple circular depressions due to degradation and superposition of impact craters. Although the existing method for detecting the impact pits based on the terrain analysis considers the terrain information of the impact pits to a certain extent, the existing method does not consider the space structure information of the impact pits, so that the extraction result precision is low.
Disclosure of Invention
In order to solve the problems, the invention provides a moon impact pit automatic extraction method and a moon impact pit automatic extraction device based on machine learning, and relates to a novel impact pit automatic extraction method combining grid-oriented and object-oriented.
A moon impact pit automatic extraction method based on machine learning comprises the following steps: acquiring DEM data of an area to be identified on the moon; identifying the area to be identified according to a pre-constructed grid-level classifier and DEM data of the area to be identified to obtain candidate impact pits; obtaining a radial elevation profile of the candidate impact pit according to the candidate impact pit and the DEM data of the area to be identified; and identifying the candidate meteorite crater according to a pre-constructed object-level classifier and the radial elevation profile of the candidate impact crater to obtain the impact crater.
In the automatic moon impact pit extraction method as described above, optionally, the identifying the region to be identified according to a pre-constructed grid-level classifier and DEM data of the region to be identified to obtain a candidate impact pit includes: according to the DEM data of the area to be recognized, determining the terrain element type of each grid in the area to be recognized; classifying each grid of the area to be identified based on a pre-constructed grid-level classifier and the terrain element type of each grid in the area to be identified to obtain a collision pit candidate grid; and obtaining the candidate impact pit according to the impact pit candidate grid.
In the automatic moon impact pit extraction method as described above, optionally, the obtaining the candidate impact pit according to the impact pit candidate grid includes: clustering the impact pit candidate grids by using a clustering algorithm to obtain clustered objects; and creating a circumscribed circle based on the clustered object to obtain the candidate impact pit.
In the automatic moon impact pit extraction method as described above, optionally, the clustering algorithm applies spatial clustering for density-based noise.
In the automatic moon impact crater extraction method as described above, optionally, the candidate meteor crater is identified based on a pre-constructed object-level classifier and the radial elevation profile of the candidate impact crater, and the number of the radial elevation profile of the candidate impact crater used in an impact crater is 12, and the radial elevation profiles are spaced by 30 °.
Another aspect provides a moon impact pit automatic extraction device based on machine learning, which includes: the acquisition module is used for acquiring DEM data of an area to be identified on the moon; the first obtaining module is used for identifying the area to be identified according to a pre-constructed grid-level classifier and DEM data of the area to be identified to obtain candidate impact pits; the second obtaining module is used for obtaining a radial elevation profile of the candidate impact pit according to the candidate impact pit and the DEM data of the area to be identified; and the third obtaining module is used for identifying the candidate meteorite crater according to a pre-constructed object-level classifier and the radial elevation profile of the candidate crater to obtain the crater.
In the automatic moon pit impact extraction apparatus as described above, optionally, the first obtaining module includes: the determining unit is used for determining the terrain element type of each grid in the area to be identified according to the DEM data of the area to be identified; the first obtaining unit is used for classifying each grid of the area to be identified based on a pre-constructed grid-level classifier and the terrain element type of each grid in the area to be identified to obtain a collision pit candidate grid; and the second obtaining unit is used for obtaining the candidate impact pit according to the impact pit candidate grid.
In the automatic moon pit extracting apparatus as described above, optionally, the second obtaining unit includes: the first obtaining subunit is used for clustering the impact pit candidate grids by using a clustering algorithm to obtain clustered objects; and the second obtaining subunit is used for creating a circumscribed circle based on the clustered object to obtain the candidate impact pit.
In the automatic moon pit extraction apparatus as described above, optionally, the clustering algorithm used by the first deriving subunit applies spatial clustering for density-based noise.
In the automatic moon impact pit extraction device as described above, optionally, the number of the radial elevation profile of the candidate impact pit used by the third obtaining module is 12, and the candidate impact pits are spaced by 30 °.
The technical scheme provided by the embodiment of the invention has the following beneficial effects:
the method comprises the steps of obtaining DEM data of an area to be recognized on the moon, recognizing the area to be recognized according to a pre-constructed grid-level classifier and the DEM data of the area to be recognized to obtain candidate impact pits, obtaining a radial elevation profile of the candidate impact pits according to the candidate impact pits and the DEM data of the area to be recognized, recognizing the candidate meteor pits according to a pre-constructed object-level classifier and the radial elevation profile of the candidate impact pits to obtain the impact pits, combining a machine learning method with geomorphic elements, and using a method of combining grid-oriented and object-oriented in machine learning, so that the recognition accuracy of the impact pits is improved.
Drawings
Fig. 1 is a schematic flow chart of a method for automatically extracting moon impact pits based on machine learning according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a crash pit database provided in the prior art with a diameter greater than 20 km;
fig. 3 is a schematic structural diagram of a training area based on fig. 2 according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a test area based on fig. 2 according to an embodiment of the present invention;
fig. 5 is a schematic diagram of a distribution of topographic elements based on the different analysis scales of fig. 3 according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of a training process of a grid-level classifier according to an embodiment of the present invention;
FIG. 7 is a schematic view in elevation in radial direction of a cross-section taken through an exemplary embodiment of the present invention;
FIG. 8 is a schematic view of a normalized radial elevation profile provided by an embodiment of the present invention;
FIG. 9 is a schematic structural diagram of a positive sample and a negative sample of a training area according to an embodiment of the present invention;
fig. 10 is a schematic flow chart of another automatic moon impact pit extraction based on machine learning according to an embodiment of the present invention;
fig. 11 is a schematic structural diagram of an automatic moon impact pit extraction device based on machine learning according to an embodiment of the present invention.
Detailed Description
The present invention will be described in detail below with reference to the embodiments with reference to the attached drawings.
Referring to fig. 1, an embodiment of the present invention provides a method for automatically extracting a moon impact pit based on machine learning, which includes the following steps:
The embodiment does not limit the manufacturing mode of the acquired lunar DEM data, can utilize the height measurement data of the laser altimeter carried by the Chang 'e satellite I to be manufactured, and can also combine the image data transmitted back by the Chang' e satellite CCD three-line-array camera with the satellite-borne laser altimeter to be manufactured. In application, the digital elevation data may be DEM data with a resolution of 500m of Chang E I, and a vertical accuracy of about 60m. And after the DEM data is produced, selecting an area to be identified.
And 102, identifying the area to be identified according to a pre-constructed grid-level classifier and DEM data of the area to be identified to obtain candidate impact pits.
Specifically, extracting the landform element type of the EDM data of the region to be identified to obtain the landform element type of the region to be identified, identifying the region to be identified by the constructed grid-level classifier according to the landform element type of the region to be identified, namely, using the landform element type of the region to be identified as the input of the constructed grid-level classifier to obtain an impact pit candidate grid and a non-impact pit grid, performing spatial clustering on the impact pit candidate grid, and creating a minimum circumcircle to obtain a candidate impact pit (or a candidate impact pit object). The Spatial Clustering algorithm may be a DBSCAN algorithm (Density-Based Clustering with Noise), and in other embodiments, other Spatial Clustering algorithms may also be used, which is not limited in this embodiment. The collision pit candidate grids and the non-collision pit grids are both image grids, and due to the fact that the number of collision pits on the lunar surface is large, the space structure is complex, the obtained binary image is high in noise, and therefore noise removing processing is conducted on the collision pit candidate grids through open operation before spatial clustering is conducted.
The construction process of the classifier is explained as follows:
the data base of the classifier is constructed, and specifically, an existing impact pit database, for example, an existing impact pit database with a diameter greater than 20km, can be selected, as shown in fig. 2. The impact pit database adopts a manual visual interpretation method, and the accuracy is high. The digital elevation data used for extracting the spatial structure information of the impact pits can adopt DEM data with ChangE I resolution of 500m, and the vertical precision is about 60m. A training area and a test area, both located in the mid latitude area on the back of the moon with more impact pits coverage, were selected in the existing impact pit database, where there were 18 and 92 impact pits in the training area and the test area, respectively, as shown in fig. 3. In fig. 3, the left diagram illustrates the training area and the right diagram illustrates the test area.
The training method of the classifier selects RF (Random Forest), which is a widely used machine learning classification method and has been successfully applied to various fields such as geomorphology and geology. During the training process, each tree in the RF is based on randomly selected input features and samples are randomly selected from the training set. For unlabeled data, the RF-based classification results are an aggregation of the classification results for each tree, weighted by their probability estimates. Overfitting during learning of complex relationships is avoided, and the method works well for relatively sparse and unbalanced data and noise. Therefore, the grid-level classifier may be referred to as a grid-level random forest classifier, and the object-level classifier described below may be referred to as an object-level random forest classifier.
The impact pits are characterized by lower, wider, flatter interior and higher, narrower, steeper edges, as shown in fig. 4, although there are impact pits with a central bulge in the interior. With respect to the terrain attributes (i.e. slope, curvature, etc.), the type of the topographic elements (topographic form elements or topographic elements) of a location may comprehensively reflect the spatial structure of the terrain surrounding the location on an analysis scale. The type of the geomorphic element is spatial structure information, which may include: 10 kinds of mountain peaks, ridge ridges, flat areas, valleys, slope shoulders, slope feet, straight back slopes, convex back slopes, concave back slopes and hollow lands. The method for extracting the types of the geomorphic elements mainly comprises the steps of carrying out space division on the earth surface form by utilizing raster DEM data and specifying the type of the belonging geomorphic element. The extraction method can adopt the existing technology, such as a multi-scale extraction method, and can be a terrain element classification method for comprehensive judgment under multiple analysis scales, and the extracted terrain element types are called multi-scale terrain element types under the method.
In the embodiment, multi-scale topographic elements are adopted to describe the spatial structure information of the grids in the impact pit, and the features (namely the multi-scale topographic elements) are input into a classifier for training. Since the topographic elements are unstable on a small analysis scale, a topographic element type determination method on a preset analysis scale is adopted. And determining a characteristic point on each preset analysis scale, wherein the characteristic point refers to one of eight field directions, and the pixel with the maximum elevation deviation between the elevation section of the unit of interest and the straight line elevation line from the end point of the analysis scale.
Because 96.2% of the impact pits on the moon have a diameter less than 60km, the 60km maximum analysis scale can effectively cover the impact pits for the grids in the impact pits, and more information of impact pits with longer distances can be avoided. The selected analysis scale is from 3km, the step length is 1km, namely two pixels are analyzed to 60km, and the loss of the impact pit space structure can be reduced to a certain extent while the calculated amount is reduced. Fig. 5 is a topographic feature distribution map on a preset analysis scale of the training area, where the analysis scale of fig. 5A is 20km, the analysis scale of fig. 5B is 40km, and the analysis scale of fig. 5C is 60km. As can be seen, as the scale of analysis increases, more grids within the impact pit are identified as topographical elements with negative topography, while the impact pit edges are identified as topographical elements with positive topography, a feature consistent with the macro-structure of the impact pit. That is, the multi-scale analysis can quantify the spatial structure of the impact pits, so that the impact pits can be more efficiently extracted.
At the grid level, the terrain that strikes the grid inside the pit and its surrounding area typically shows variations in convexity and concavity at different analysis scales (i.e., different window sizes). Thus, the input features of each sample of a grid-level classifier (or grid-level classifier) are multi-scale features of the sample unit over a series of successive analysis scales. Positive and negative samples of a grid level of the impact pit are determined and recognized according to an existing impact pit database, DEM data (namely multi-scale landform elements) with impact pit space structure information are used as input features of a classifier, and the grid level classifier is obtained through training. A positive sample indicates impact pits in the sample and a negative sample indicates no impact pits in the sample.
After the grid-level classifier is trained in the training area, the classifier is applied to the testing area to verify the accuracy of the classifier. And when the accuracy requirement is met, finishing the training of the grid-level classifier, otherwise, adjusting the parameters of the grid-level classifier until the accuracy requirement is met in the test area.
And after the grid-level classifier is constructed, performing geomorphic element type extraction on the EDM data of the area to be identified to obtain the geomorphic element type of the area to be identified. The grids may be classified into bump candidate grids or non-bump grids according to their types of topographical elements. The partitioning is done by a grid-level classifier. And identifying the area to be identified by using the constructed grid-level classifier according to the landform element type of the area to be identified, namely using the landform element type of the area to be identified as the input of the constructed grid-level classifier to obtain an impact pit candidate grid and a non-impact pit candidate grid, and then converting the impact pit candidate grid into a candidate impact pit. The conversion process comprises the following steps: remove outliers (e.g., regions smaller than a first preset threshold, or regions at a distance from other candidate grids greater than a second preset threshold), and then: removing noise by using morphological open operation in mathematics, then creating impact pit candidate grids by using a DBSCAN algorithm, namely, clustering the grids based on the density of the grids to obtain grid cluster clusters in any shape, and then creating a minimum circumscribed circle to obtain candidate impact pits, namely impact pit candidate objects. Referring to FIG. 6, a process for training a grid-level classifier is illustrated.
It should be noted that: impact pits formed by impact are basically nearly circular, and although the shapes of the impact pits are changed due to the overlapping and the degradation of the impact pits at the later stage, the impact pits are theoretically circular, so that a circumscribed circle is selected when the impact pits are created.
And 103, obtaining a radial elevation profile of the candidate impact pit according to the candidate impact pit and the DEM data of the area to be identified.
In step 102, the resulting candidate impact pits are only a possible aggregation of impact pit grid cells, and the radial elevation profile is considered for determining whether a true impact pit is present. The elevation profile may to some extent well reflect the spatial structure of the impact pit. Although the impact pit is theoretically a circular recessed region, in practice, the spatial structure of the impact pit is more complicated due to degradation and superposition of the impact pit. Therefore, the elevation profiles of the impact pits in different directions can effectively characterize their detailed spatial structural features. For a single candidate impingement pit, multiple radial elevation profiles, e.g., 12, etc., may be selected.
And 104, identifying the candidate meteorite crater according to a pre-constructed object-level classifier and a radial elevation profile of the candidate crater to obtain the crater.
Fig. 8 is a radial elevation profile of an impact pit, and if the input feature is the radial elevation profile, it may be determined whether the candidate impact pit is an impact pit, and in order to improve the recognition accuracy, for the candidate impact pit, the input feature is a plurality of radial elevation profiles, for example, 12 radial elevation profiles, each adjacent radial elevation profile is spaced by 30 degrees, and it is determined whether each radial elevation profile is an impact pit, and if it is determined that the number of impact pits is greater than or equal to a preset number threshold, the candidate impact pit is an impact pit, otherwise, the candidate impact pit is a non-impact pit. The preset number threshold is a positive number, which may be half of the plurality of pieces, or 0.75 times of the plurality of pieces, and the specific numerical value of the preset number threshold is not limited in this embodiment.
The following explains the construction process of the object-level classifier:
the process mainly comprises two parts: training samples and spatial structure information input features at the object level are collected. The radial elevation profile of the impact pit is shown in fig. 7, and by taking the elevation profile of the impact pit as a training sample of the impact pit, the elevation profiles in different directions can effectively show the detailed spatial structural characteristics of the impact pit.
Typically, elevation profiles for different impingement pits will have different lengths and cushions, and therefore, normalization is performed on each elevation profile, as shown in FIG. 8, to convert the average elevation of each portion of the radial elevation profile to between 0 and 1, with 0 and 1 corresponding to the minimum and maximum elevation values, respectively, of the profile. The input feature for each sample of the object-level classifier is a normalized elevation value of medium length from the sample section [0,1].
To train the subject-level random forest classifier, 12 radial elevation profiles of a single impact pit in the training area were collected as positive samples, all angles between adjacent profiles were 30 °, as shown in fig. 9, with the abscissa in fig. 9 representing the lateral division of the training area into 9 segments, each segment being one length unit. Each impact pit in the training area may provide a number of elevation profiles. Even if the training area has few impact pits, a large number of positive samples can be collected at the target level to train the object-level random forest classifier. For each impact pit in the training area, a circle of similar size is randomly generated in the non-impact pit area of the training data. Subsequently, 12 radial elevation profiles were collected from these circles as negative examples for training a random forest classifier construction at the subject level.
After the object-level random forest classifier is trained in the training area, the candidate impact pits obtained in step 102 are detected. For each impact pit candidate, the radial elevation profile is extracted in the same way as the positive example samples are collected at the object level. Each radial elevation profile is converted into input features of an object-level random forest classifier to evaluate whether the contour is a pit strike. If a candidate object for an impact pit is characterized by a radial elevation profile and is determined to be an impact pit, then the impact pit candidate object is the determined impact pit. Thus, the proposed method can efficiently detect those degraded and superimposed impact pits.
Referring to fig. 10, that is: the method trains a classifier (such as a random forest classifier) according to the existing impact pits and the impact pit space structure information obtained by combining digital terrain analysis, thereby realizing the automatic extraction of the impact pits. The method comprises the following steps that firstly, grids of impact pits are used as samples, multi-scale terrain element information obtained through terrain analysis is used as characteristics, a grid-level random forest classifier is trained and used for identifying candidate impact pit grids in a target area, and candidate impact pit objects are obtained; in the second stage, a random forest classifier at an object level is trained by using the radial elevation profile of the impact pits in the impact pit distribution map so as to judge whether each candidate object obtained in the first stage is the impact pit.
In general, the method combines a machine learning method with topographical elements. In the method, the random forest effectively utilizes multi-scale landform elements obtained by digital landform analysis, not only quantifies morphological information, but also quantifies the space structure inside a single impact pit, combines the space structure information obtained by the landform analysis with machine learning, improves the accuracy of a machine learning model, and expands the method to extract landform types.
Referring to fig. 11, an embodiment of the present invention provides an automatic moon impact pit extracting apparatus based on machine learning, which includes: the system comprises an acquisition module 201, a first obtaining module 202, a second obtaining module 203 and a third obtaining module 204.
Specifically, the obtaining module 201 is configured to obtain DEM data of an area to be identified on the moon. The first obtaining module 202 is configured to identify the area to be identified according to a pre-established grid-level classifier and DEM data of the area to be identified, so as to obtain a candidate impact pit. The second obtaining module 203 is configured to obtain a radial elevation profile of the candidate impact pit according to the candidate impact pit and the DEM data of the area to be identified. The third obtaining module 204 is configured to identify the candidate merle crate according to a pre-constructed object-level classifier and a radial elevation profile of the candidate crater, so as to obtain the crater.
Optionally, the first obtaining module 202 includes: the device comprises a determining unit, a first obtaining unit and a second obtaining unit. The determining unit is used for determining the type of the landform element of each grid in the area to be identified according to the DEM data of the area to be identified. The first obtaining unit is used for classifying each grid of the area to be identified based on a pre-constructed grid-level classifier and the landform element type of each grid in the area to be identified to obtain the collision pit candidate grid. The second obtaining unit is used for obtaining the candidate impact pit according to the impact pit candidate grid.
Optionally, the second obtaining unit includes: a first deriving subunit and a second deriving subunit. The first obtaining subunit is used for clustering the impact pit candidate grids by using a clustering algorithm to obtain a clustered object. And the second obtaining subunit is used for creating a circumscribed circle based on the clustered objects to obtain the candidate impact pits.
Optionally, the clustering algorithm used by the first deriving subunit applies spatial clustering for density-based noise.
Optionally, the number of radial elevation profiles of candidate impact pits used by the third deriving module 204 is 12, and are spaced 30 ° apart from each other.
It should be noted that: in the automatic moon pit extracting device provided by the above embodiment, only the division of the above functional modules is taken as an example for extraction, and in practical application, the above function distribution can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the above described functions. In addition, the automatic moon impact pit extraction device provided by the above embodiment and the automatic moon impact pit extraction method embodiment belong to the same concept, and specific implementation processes thereof are detailed in the method embodiment and are not repeated here.
It will be appreciated by those skilled in the art that the invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The embodiments disclosed above are therefore to be considered in all respects as illustrative and not restrictive. All changes which come within the scope of or equivalence to the invention are intended to be embraced therein.
Claims (10)
1. An automatic moon impact pit extraction method based on machine learning is characterized by comprising the following steps:
acquiring DEM data of an area to be identified on the moon;
identifying the area to be identified according to a pre-constructed grid-level classifier and DEM data of the area to be identified to obtain candidate impact pits;
obtaining a radial elevation profile of the candidate impact pit according to the candidate impact pit and the DEM data of the area to be identified;
and identifying the candidate meteorite crater according to a pre-constructed object-level classifier and the radial elevation profile of the candidate impact crater to obtain the impact crater.
2. The automatic moon impact pit extraction method according to claim 1, wherein the identifying the area to be identified according to a pre-constructed grid-level classifier and DEM data of the area to be identified to obtain candidate impact pits comprises:
according to the DEM data of the area to be identified, determining the type of the landform element of each grid in the area to be identified;
classifying each grid of the area to be identified based on a pre-constructed grid-level classifier and the landform element type of each grid in the area to be identified to obtain a collision pit candidate grid;
and obtaining the candidate impact pit according to the impact pit candidate grid.
3. The automatic moon impact pit extraction method according to claim 2, wherein the obtaining of the candidate impact pits from the impact pit candidate grid comprises:
clustering the impact pit candidate grids by using a clustering algorithm to obtain clustered objects;
and creating a minimum circumcircle based on the clustered objects to obtain the candidate impact pits.
4. The method for automatically extracting moon impact pits according to claim 3, wherein the clustering algorithm is a density-based noise application spatial clustering algorithm.
5. The method for automatically extracting moon impact craters according to claim 1, wherein the candidate merle crates are identified based on a pre-constructed object-level classifier and the radial elevation profiles of the candidate impact crates, and the number of the radial elevation profiles of the candidate impact crates used in the impact crates is 12, and the profiles are spaced by 30 °.
6. An automatic moon impact pit extraction device based on machine learning, characterized in that the automatic moon impact pit extraction device comprises:
the acquisition module is used for acquiring DEM data of an area to be identified on the moon;
the first obtaining module is used for identifying the area to be identified according to a pre-constructed grid-level classifier and DEM data of the area to be identified to obtain candidate impact pits;
the second obtaining module is used for obtaining a radial elevation profile of the candidate impact pit according to the candidate impact pit and DEM data of the area to be identified;
and the third obtaining module is used for identifying the candidate meteorite crater according to a pre-constructed object-level classifier and the radial elevation profile of the candidate crater to obtain the crater.
7. The automatic moon impact pit extraction device according to claim 6, wherein the first obtaining module comprises:
the determining unit is used for determining the type of the landform element of each grid in the area to be identified according to the DEM data of the area to be identified;
the first obtaining unit is used for classifying each grid of the area to be identified based on a pre-constructed grid-level classifier and the landform element type of each grid in the area to be identified to obtain a collision pit candidate grid;
and the second obtaining unit is used for obtaining the candidate impact pit according to the impact pit candidate grid.
8. The automatic moon pit extraction apparatus according to claim 7, wherein the second obtaining unit includes:
the first obtaining subunit is used for clustering the impact pit candidate grids by using a clustering algorithm to obtain clustered objects;
and the second obtaining subunit is used for creating a circumscribed circle based on the clustered object to obtain the candidate impact pit.
9. The automatic moon impact pit extraction device according to claim 8, wherein the clustering algorithm used by the first deriving subunit applies spatial clustering for density-based noise.
10. The automatic moon impact pit extraction device according to claim 8, wherein the number of radial elevation profile of the candidate impact pits used by the third obtaining module is 12, and the candidate impact pits are spaced 30 ° apart from each other.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210956296.9A CN115272769A (en) | 2022-08-10 | 2022-08-10 | Automatic moon impact pit extraction method and device based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210956296.9A CN115272769A (en) | 2022-08-10 | 2022-08-10 | Automatic moon impact pit extraction method and device based on machine learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115272769A true CN115272769A (en) | 2022-11-01 |
Family
ID=83751593
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210956296.9A Pending CN115272769A (en) | 2022-08-10 | 2022-08-10 | Automatic moon impact pit extraction method and device based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115272769A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107818147A (en) * | 2017-10-19 | 2018-03-20 | 大连大学 | Distributed temporal index system based on Voronoi diagram |
| CN108305301A (en) * | 2018-01-25 | 2018-07-20 | 山东大学 | A kind of menology multi-dimension texture generation method based on remote sensing image |
| CN110334645A (en) * | 2019-07-02 | 2019-10-15 | 华东交通大学 | A kind of moon impact crater recognition methods based on deep learning |
| CN111721302A (en) * | 2020-06-24 | 2020-09-29 | 北京理工大学 | Method for recognizing and sensing complex terrain features on surface of irregular asteroid |
| WO2021125395A1 (en) * | 2019-12-18 | 2021-06-24 | 한국항공우주연구원 | Method for determining specific area for optical navigation on basis of artificial neural network, on-board map generation device, and method for determining direction of lander |
| US20220108116A1 (en) * | 2020-10-06 | 2022-04-07 | Wipro Limited | Method and system for navigating vehicles based on road conditions determined in real-time |
| CN114386300A (en) * | 2021-12-28 | 2022-04-22 | 吉林大学 | Full-moon landing area addressing method based on evidence weight and fractal |
-
2022
- 2022-08-10 CN CN202210956296.9A patent/CN115272769A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107818147A (en) * | 2017-10-19 | 2018-03-20 | 大连大学 | Distributed temporal index system based on Voronoi diagram |
| CN108305301A (en) * | 2018-01-25 | 2018-07-20 | 山东大学 | A kind of menology multi-dimension texture generation method based on remote sensing image |
| CN110334645A (en) * | 2019-07-02 | 2019-10-15 | 华东交通大学 | A kind of moon impact crater recognition methods based on deep learning |
| WO2021125395A1 (en) * | 2019-12-18 | 2021-06-24 | 한국항공우주연구원 | Method for determining specific area for optical navigation on basis of artificial neural network, on-board map generation device, and method for determining direction of lander |
| CN111721302A (en) * | 2020-06-24 | 2020-09-29 | 北京理工大学 | Method for recognizing and sensing complex terrain features on surface of irregular asteroid |
| US20220108116A1 (en) * | 2020-10-06 | 2022-04-07 | Wipro Limited | Method and system for navigating vehicles based on road conditions determined in real-time |
| CN114386300A (en) * | 2021-12-28 | 2022-04-22 | 吉林大学 | Full-moon landing area addressing method based on evidence weight and fractal |
Non-Patent Citations (1)
| Title |
|---|
| YAN-WEN WANG: ""Automatic Crater Detection by Training Random Forest Classifiers with Legacy Crater Map and Spatial Structural Information Derived from Digital Terrain Analysis"" * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112287807B (en) | Remote sensing image road extraction method based on multi-branch pyramid neural network | |
| CN102496034B (en) | High-spatial resolution remote-sensing image bag-of-word classification method based on linear words | |
| CN108830870B (en) | Satellite image high-precision farmland boundary extraction method based on multi-scale structure learning | |
| CN111028255A (en) | Farmland area pre-screening method and device based on prior information and deep learning | |
| CN104951799A (en) | SAR remote-sensing image oil spilling detection and identification method | |
| CN112099046B (en) | Airborne LIDAR three-dimensional plane detection method based on multi-value voxel model | |
| CN102214290B (en) | License plate positioning method and license plate positioning template training method | |
| US20220414376A1 (en) | Automated computer system and method of road network extraction from remote sensing images using vehicle motion detection to seed spectral classification | |
| CN106952242A (en) | A kind of progressive TIN point cloud filtering method based on voxel | |
| CN113360587B (en) | Land surveying and mapping equipment and method based on GIS technology | |
| CN110794424A (en) | Full-waveform airborne laser radar ground feature classification method and system based on feature selection | |
| CN107832849A (en) | The power line gallery 3-D information fetching method and device in a kind of knowledge based storehouse | |
| Özdemir et al. | Automatic extraction of trees by using multiple return properties of the lidar point cloud | |
| CN116168246A (en) | Method, device, equipment and medium for identifying waste slag field for railway engineering | |
| CN117197677A (en) | Tropical rain forest arbor-shrub separation method based on laser radar point cloud data | |
| CN116415843A (en) | Multi-mode remote sensing auxiliary mine ecological environment evaluation method for weak network environment | |
| CN116823896A (en) | Target mining area range prediction method and device under high vegetation coverage and electronic equipment | |
| CN118247662A (en) | Environment detection method for mining | |
| CN113219472B (en) | Ranging system and method | |
| CN112488190A (en) | Point cloud data classification method and system based on deep learning | |
| CN117078925A (en) | Building rubbish annual output accurate calculation method based on RDSA-deep LabV3+ network | |
| CN117115671A (en) | Soil quality analysis method and device based on remote sensing and electronic equipment | |
| CN114743059B (en) | Automatic classification method for submarine geographic entities by integrating topographic features | |
| CN104112145A (en) | Facial expression recognition method based on PWLD and D-S evidence theory | |
| CN115272769A (en) | Automatic moon impact pit extraction method and device based on machine learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20221101 |