CN109816707A - A kind of field of opencast mining information extracting method based on high-resolution satellite image - Google Patents
A kind of field of opencast mining information extracting method based on high-resolution satellite image Download PDFInfo
- Publication number
- CN109816707A CN109816707A CN201811589599.1A CN201811589599A CN109816707A CN 109816707 A CN109816707 A CN 109816707A CN 201811589599 A CN201811589599 A CN 201811589599A CN 109816707 A CN109816707 A CN 109816707A
- Authority
- CN
- China
- Prior art keywords
- satellite image
- field
- opencast mining
- image
- information extracting
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The present invention relates to Remote Sensing Image Processing Technology fields, in particular to a kind of field of opencast mining information extracting method based on high-resolution satellite image.The following steps are included: 1), to satellite image pre-process；2) the optimum segmentation scale for, calculating field of opencast mining and other regional background environmental elements, constructs the segmentation hierarchical structure of satellite image；3) the characterization factor collection of each cutting object, is calculated, layering exposure mask removes regional background environmental element related with characterization factor collection；4) decision tree, is constructed by random forests algorithm, is classified according to characterization factor collection to the remaining element that mixes after eliminating regional background environmental element related with characterization factor collection, field of opencast mining information is obtained.The present invention using random forest sorting algorithm, can fast and accurately obtain the information of field of opencast mining, processing mode is simple, efficient by carrying out Hierarchical Segmentation processing to satellite image.
Description
Technical field
The present invention relates to Remote Sensing Image Processing Technology fields, in particular to a kind of based on the outdoor of high-resolution satellite image
Stope information extracting method.
Background technique
With remote sensing satellite cause is greatly developed in recent years, the High Resolution Remote Sensing Satellites that China has possessed autonomous property right are high
Dividing No.1 (panchromatic image resolution ratio is 2m, multispectral resolution rate 10m) and high score two, (panchromatic image resolution ratio is 0.8m, more
Spectral resolution 2m), however for the field of opencast mining of regional scale, target knowledge is carried out with high-resolution remote sensing image at present
Method for distinguishing is mainly still based on artificial visual interpretation.Being greatly improved for satellite-remote-sensing image spatial resolution, can be more
More careful atural object element information abundant is clearly showed, area can be improved in remote sensing and the combination of object-oriented image processing algorithm
The information extraction efficiency and accuracy of field of opencast mining under the scale of domain.
The core of field of opencast mining information extracting method based on domestic high resolution ratio satellite remote-sensing image is: 1. domestic
High resolution ratio satellite remote-sensing image pretreatment；2. the foundation of Image Segmentation hierarchical structure；3. remote sensing and machine learning algorithm are revealing
Combination in its stope information extraction.
By the long-term and unremitting painstaking research of Chinese scholar, in middle low resolution remote sensing image, (spatial resolution is greater than
On the basis of research achievement 15m), the relevant research of field of opencast mining information extraction is further deepened, is refined, and achieves
Original research achievement.2004, the researchers such as WANG Xiaohong, Nie Hongfeng were research area with Jiangxi Province Chongyi County, were based on
Two kinds of High Resolution Remote Sensing Satellites images of QuickBird image and Spot-5 have carried out south using the method for visual indoor interpretation
In the high vegetative coverage mountain area in side, extensive mining region mining information extraction, illegal mining under the conditions of complicated field investigation
The correlative studys such as movement monitoring and mine development condition evaluation, compared to low resolution remote sensing image in use in the past, interpretation effect
Fruit is more accurate, and further saving monitoring and surveying cost, (high resolution satellite remote sensing image is in mine development situation and environmental monitoring
Application effect compare, land resources remote sensing, the 16th phase in 2004).Currently, the High Resolution Remote Sensing Satellites of China's independent research
Including two series: resource series satellite and high score series of satellites.2016, the small researchers such as graceful of Xiong Qianjin, bavin used state
No. 3 satellite high-resolution remote sensing images of resource are produced, mining state has been determined by the calculating of optimal bands combined index value, has been opened
Adopt range (data processing method of No. 3 satellite images of resource in Mine Monitoring application, Wuhan Iron and Steel Plant technology, the 54th phase in 2016)；
And the researchers such as Wei Jianglong, Zhou Yingzhi are based on domestic high score No.1 remote sensing image can manage Polymetallic Ore Deposit as research area
Data source by data prediction, establishes the research work such as remote Sensing Interpretation mark, statistical analysis, is extracted mine in research area
Distribution and mining information (the mine remote sensing monitoring based on one number of high score --- for it can manage Polymetallic Ore Deposit, there is coloured gold
Belong to: mine part, the 68th phase in 2016)；On this basis, the researchers such as Ma Xiuqiang, Peng's order also attempt two number of high score
According to applied in the investigation of Hubei Province Daye mine geological environment, they pass through analysis earth's surface spectral characteristic of ground and space characteristics,
Using band math and density slice, qualitative analysis mining area water pollution rank, comprehensive analysis Diggings soil utilization obstacle, and it is right
The mine development situation in the important area Kuang Ji and environmental change are monitored that (two number of high score is in Daye, hubei Province mine geological environment
Application in investigation, land resources remote sensing, the 29th phase in 2017).
In addition to traditional artificial visual decomposition method, into after the high-resolution remote sensing image epoch, still have many
Researcher is continuing supervised classification and unsupervised classification in the research of mine remote sensing investigation and the application aspect in monitoring, such as
Chen Xingjie used domestic GF-1 number in 2017, and it is several to compared maximum likelihood method, minimum distance method, support vector machines etc.
Application effect of the supervised classification method in terms of land use and covering, and supervised classification method is discussed in high score image
Using various aspects difference (supervised classification method based on GF-1 satellite image compares, mining survey, 2017 the 45th
Phase)；In addition, Ikonos image was used in 2013 for researchers such as Jingjing, Wang Denghong, with ISODATA non_monitor algorithm
Assessed Rare Earth Mine limit of mining and periphery river pollution degree that (IKONOS remotely-sensed data is in ion adsorption type rare earth ore area
Application study in environmental pollution surveys --- by taking the area of south jiangxi Xunwu as an example, earth journal, the 3rd phase in 2013).
Traditional remote sensing image interpretation method substantially still understands remote sensing image from angle pixel-based, this
The characteristics of method can only reflect the spectral signature of single pixel, can not understand image on the whole does not utilize the single of image
Connection between characteristics of objects and object, for high-resolution remote sensing image, traditional method pixel-based will lead to shadow
As the bulk redundancy of information and the waste of resource.
Summary of the invention
Present invention aim to solve above-mentioned background technique, there are the presence of traditional remote sensing image decomposition method will lead to
The problem of waste of the bulk redundancy of image information and resource, provide a kind of surface mining based on high-resolution satellite image
Field information extracting method.
The technical solution of the present invention is as follows: a kind of field of opencast mining information extracting method based on high-resolution satellite image,
It is characterized by comprising following steps: 1), to shooting having the domestic high-resolution satellite image of target field of opencast mining to carry out pre-
Processing；
2) it, calculates field of opencast mining and appears in the optimum segmentation ruler of other regional background environmental elements in satellite image
Degree, constructs the segmentation hierarchical structure of satellite image；
3), based on pretreated satellite image is passed through, the characterization factor collection of each cutting object is calculated, layering exposure mask is gone
Except regional background environmental element related with characterization factor collection；
4) decision tree, is constructed by random forests algorithm, it is related with characterization factor collection to eliminating according to characterization factor collection
Regional background environmental element after remaining mixing element classify, obtain field of opencast mining information.
In the further step 1), pretreated method is carried out to satellite image are as follows: have target outdoor shooting
The domestic high-resolution satellite image of stope carries out the processing that radiation calibration, atmospheric correction, ortho-rectification, image co-registration are handled
Operation.
In the further step 2), the method for calculating optimum segmentation scale are as follows: use two index method of Euclid
The optimum segmentation scale for carrying out the various geological environment elements of rational judgment, for that can not determine area by two index method of Euclid
The geological environment element divided, divides object unit by the method for Image Segmentation.
In the step 2), the method that constructs the segmentation hierarchical structure of satellite image are as follows: reference segmentation ruler
Angle value extracts the principle construction information extraction layer of geological environment element from the easier to the more advanced, by using to different geology environmental elements
Different segmentation scales determines the segmentation scale for being suitble to every kind of geological environment element, establishes reliable classification gauge on this basis
Then collect, all geological environment elements that can not be segmented are set as the bottom.
In the further step 3), the characterization factor collection includes spectral signature, geometrical characteristic, textural characteristics
And space characteristics.
The further spectral signature includes each wave band average gray, each wave band gray standard deviation, brightness value, returns
One changes vegetation index and maximum difference.
The geometrical characteristic includes length-width ratio, density, compact degree, shape index, rectangle fitting degree, not right
Title property, circularity.
The textural characteristics include homogeney based on gray level co-occurrence matrixes, contrast, distinctiveness ratio, entropy,
Mean value, standard deviation, similarity and the entropy based on grey scale difference vector, mean value, contrast.
The further space characteristics include away from industrial and mineral warehouse land distance and neighborhood contact relation.
In the further step 4), pass through the method for random forests algorithm building decision tree are as follows: using random choosing
The method for taking Split Attribute collection selects the training sample set of 2/3 size, 1/3 remaining data from entire training center data set
The learning ability of random forests algorithm is estimated as data outside bag, is repeated this process ntree times, 300≤ntree≤500,
Ballot finally, which is carried out, using the prediction result of ntree decision tree of generation selects final classification results.
Advantages of the present invention has: 1, comparing traditional indoors artificial visual interpretation, this technical solution provides one kind and exists
The method of field of opencast mining is identified on big regional scale, while improving information extraction efficiency to a certain extent；
2, by Image Segmentation hierarchical structure, the Multi-layer technology to all kinds of geological environment elements is met, while being obtained every
One layer of characterization factor rule set is spy of the quantitative analysis difference atural object element on domestic high resolution ratio satellite remote-sensing image
Sign rule provides foundation；
3, due to having used random forest sorting algorithm, this technical solution can satisfy the domestic high-resolution of big data quantity
The accurate extraction of field of opencast mining on satellite remote-sensing image, this is that other methods are also unconsummated.
The present invention is by carrying out Hierarchical Segmentation processing to satellite image, using random forest sorting algorithm, can quickly and
The accurate information for obtaining field of opencast mining, processing mode is simple, efficient, and solving existing basis, there are information redundancies and resource wave
The problem of taking there is great promotional value.
Detailed description of the invention
Fig. 1: field of opencast mining information extracting method flow chart；
Fig. 2: remote sensing image pretreatment process；
Fig. 3: all kinds of geological environment element sample for reference figures；
Fig. 4: 6 ground object information extraction hierarchical charts that the present embodiment is established；
Specific embodiment
The following further describes the present invention in detail with reference to the accompanying drawings and specific embodiments.
As shown in Figure 1, the field of opencast mining information based on domestic high resolution ratio satellite remote-sensing image for the present embodiment mentions
Take method flow diagram.Present embodiments provide a kind of field of opencast mining information extracting method, comprising the following steps:
Step 1: obtaining the original domestic high-resolution remote sensing image of target area, and pre-processed.
Obtain the original domestic high-resolution remote sensing image of target area.In specific implementation, original high resolution remote sensing shadow
As No. two satellite high-resolution remote sensing images of high score No.1 satellite or high score can be selected, their multi light spectrum hands all contains 4
The information (respectively blue, green, red, near infrared band) of wave band, specific payload technical indicator is as shown in table 1-2:
1 high score No.1 Satellite Payloads technical indicator of table
Table No. two Satellite Payloads technical indicators of 2 high score
The design embodiment has chosen the Luoyang City Yu Xi Luanchuan Molybdenum openpit and exploits mining area in October, 01 to 2017 in 2016
High score No.1 satellite remote-sensing image, for as the high-definition remote sensing image data to target area field of opencast mining.Tool
During body is implemented, which further includes domestic high-resolution remote sensing image pretreatment, usually need to be by pre- place after remote sensing image obtains
It could be used as the data source of information extraction after reason, as shown in Fig. 2, domestic high-resolution remote sensing image pretreatment process includes
The operation such as radiation calibration, atmospheric correction, ortho-rectification and image co-registration.
Step 2: determining optimum segmentation scale.Quantitatively sentence in this case study using two index method of Euclid
Break the optimum segmentation scales of various atural objects, the specific implementation process is as follows.
1, sample is chosen to a wide range of geologic setting element (settlement place, forest land, arable land) respectively in ArcGIS to make
For referring to polygon, and reference area size.All kinds of geological environment element sample for reference figures are as shown in Figure 3:
2, four wave band weights of GF-1 image are 1 after the fusion of this experimental setup, when forest land segmentation range scale is by 50
To 450, settlement place and arable land segmentation range scale are by 50 to 250；Compact degree factor Compactness parameter is 0.5；Shape ginseng
Number Shape is less than or equal to 0.5, and value 0.1,0.3,0.5 carries out multi-scale division experiment.By the segmentation result vector file of test
It is laid out analysis and calculating referring to polygon in importing ArcGIS, and corresponding thereto, obtains 3 kinds of geological environment elements
Optimum segmentation parameter list 3:
3 various regions environmental element optimum segmentation parameter of table
Step 3: structure and information extraction rule are established layer by layer for information extraction.
Comprehensive geology environment category optimum segmentation parameter and image feature, reference segmentation scale-value, extract atural object by easily to
Difficult principle construction information extraction layer determines that every kind of atural object is suitble to certainly by using different segmentation scales to different atural object elements
Oneself segmentation scale establishes reliable classifying rules collection on this basis, the 6 ground object information extraction layers established in the present embodiment
Secondary structure will be as shown in figure 4, information extraction will carry out in the following order.
In the present embodiment from available characterization factor concentration selected brightness (Brightness), density (Density),
Gray level co-occurrence matrixes contrast (GLCMContrast), near infrared band gray average (NIRmean), rectangle fitting degree
(Rectangular fit), circularity (Roundness), shape index (Shape index), normalized differential vegetation index (NDVI)
Equal characterization factors.Using training center data set, according to the layer-by-layer selected characteristic of taxonomical hierarchy and corresponding characterization rules are established, it is special
Sign rule set is shown in Table 4.
4 characterization rules collection of table
The result of Multi-layer technology will be inherited other layer of atural object to 1 layer of Level by the correlation of class interlayer, such as
3 layers of the Level forest land that inherit in 4 layers of Level, Level 3 is father's layer of Level 4 at this time, the forest land in Level 3
It may be defined as: Class-relation to sub objects > Existence of > forest land；Similarly, at this time 4 layers of Level work
For 3 layers of sublayer of Level, the forest land in 4 layers of Level can also be by sublayer contextual definition are as follows: Class-relation to
Super objects > Existence of > forest land.Interlayer relation in this way finally obtains the remote sensing image of layering exposure mask
Figure, wherein other 4 this kind include that 4 major class (excavating plant, Tailings Dam, surface plant, bare area) are difficult to by being layered cut section
Separated geological environment element.
Using characterization rules collection, layering segmentation equally is carried out to test section data set, obtains containing road, river lake
The layering segmentation result figure of the 6 class results such as pool, settlement place, forest land, arable land and other 4.
Step 4: being wanted based on random forest sorting algorithm to other 4 kinds of geological environments on the basis of being layered segmentation result
Element is classified, and field of opencast mining is extracted.
In Henan west, research area training center carries out samples selection to other 4 kinds of geological environment elements, and this section research is selected at random
1245 sample objects, wherein 403 excavating plant samples, 292 Tailings Dam samples, 270 surface plant samples, 280
A bare area sample is concentrated from characterization factor and chooses 23 characteristic parameters participation model foundations, specific features collection parameter such as table 5.
5 feature set of table summarizes
1, spectral signature
It is the optical physics category determined by true atural object and image formation state for describing the spectral information of imaged object
Property, related to the gray value of object, it includes content and specific mathematical definition are as follows:
A, mean value (mean), the layer value C of all n pixels by constituting imaged objectLiA layer average value is calculated, it may be assumed that
Wherein: CLi--- the spectrum gray value of single pixel；
N --- pixel quantity；
B, brightness (brightness), the summation of the image bearing layer average value comprising spectral information is divided by the number of plies.To image pair
It is exactly the mean value of spectrum average as, it may be assumed that
Wherein: Ci--- the spectrum gray average of single image bearing layer；
nL--- image layer number；
C, maximum difference (max diff) calculates maximum difference, as long as the maximum value of imaged object is subtracted its minimum value
, need the maximin to all channels of object all to compare, last result is again divided by brightness.
C=max [max (CLi)-min(CLi)]/b, i=1,2,3 ..., n
Wherein: CLi--- the spectrum gray value of single pixel；
B --- brightness value of image；
N --- pixel quantity.
D, standard deviation (standard deviation), the picture of the n pixel by constituting all channels of imaged object
Standard deviation is calculated in plain value, it may be assumed that
Wherein: CLi--- the spectrum gray value of single pixel；
Ci--- the spectrum gray average of single image bearing layer；
N --- pixel quantity.
E, normalized differential vegetation index (NDVI), the principle for extracting vegetation index is difference according to reflectance,
By operations such as ratios to protrude vegetation characteristics, extract vegetation pattern or estimation green bio amount.Based on green plants leaf
Eucaryotic cell structure has high reflection near infrared band, and its chlorophyll has the physical features of strong absorbent in red spectral band, absolutely
Most vegetation index all utilizes the reflectivity information of red wave band and near infrared band, extracts vegetation relevant information.Normalization is planted
Higher by the value of index, vegetation coverage is higher；Conversely, then closer to the impervious surface without vegetation development, it may be assumed that
Wherein: NIR --- near infrared band；
R --- red wave band；
2, shape feature
It is the space geometry attribute determined by true atural object for describing the shape information of imaged object itself, packet
It is as follows containing content and mathematical definition:
A, length-width ratio (Length/Width), length-width ratio are equal to the ratio of the characteristic value of covariance matrix, biggish feature
Value is the molecule of score, it may be assumed that
R=eig1(s)/eig2(s), eig1(s) > eig2(s)
Wherein: eig1(s) and eig2 (s) be respectively covariance matrix characteristic value；
B, density (Density), density d can be expressed as imaged object area except its upper radius, be retouched using density
State the compactness of imaged object.Ideal compact shape is a square, an image pair in the figure of grids of pixels
For the shape of elephant closer to square, its density is higher, and n is the pixel quantity for constituting imaged object here, and radius is using association
Variance matrix carrys out approximate calculation, it may be assumed that
Wherein: n --- pixel quantity；
C, compact degree (Compactness), the long k of object is multiplied by wide q, then divided by the number of pixels n of object, i.e.,；
P=k × q/n
Wherein: p --- compact degree；
The length of k --- object；
The width of q --- object；
The number of pixels of n --- object；
D, shape index (shape index), mathematically shape index is the boundary length of imaged object except its upper face
Long-pending subduplicate 4 times.The smoothness on imaged object boundary can be described using shape index S, imaged object is more broken, then it
Shape index it is bigger, it may be assumed that
Wherein: S --- the smoothness on imaged object boundary can be described；
The boundary length of e --- imaged object；
The area of A --- imaged object；
E, rectangle fitting degree (Rectangular Fit), the first step for calculating rectangle adaptability are to generate one and consider
Object rectangle of the same area considers the length-width ratio of object in calculating process, hereafter, compare object area outside rectangle and
In rectangle not by object filling area ratio, if value is 0, it is meant that this rectangle is improper；If value is 1, explanation
It can be well suited for object；
F, asymmetry (Asymmetry), an imaged object is longer, its asymmetry is higher, for an image
For object, can an approximate ellipse, asymmetry is represented by the length ratio of elliptical short axle s and long axis l, with asymmetry
Property increase and characteristic value increase, it may be assumed that
K=1-s/l
Wherein: k --- asymmetry；
S --- imaged object minor axis length；
L --- imaged object long axis length；
G, circularity (Roundness) calculates cutting unit close to circular degree, and the difference of maximum radius and least radius is
When 0, circularity 0, the bigger expression shape of circularity more deviates circle,
R=Wmax-Wmin
Wherein: Wmax--- cutting unit maximum radius；
Wmin--- cutting unit least radius；
3, textural characteristics
Texture occupies more important position in high-resolution remote sensing image has divided, and exists for describing the spectrum of imaged object
Rule in spatial distribution reflects certain variation of the color, geometry and gray scale on atural object surface, and what is most generally used retouches
The method for stating texture has gray level co-occurrence matrixes (Grey Level Concurrence Matrix, GLCM) and grey scale difference vector
Matrix (Grey Level difference vector, GLDV), on this basis using more texture statistics amount include with
Under several classes:
A, homogenieity (Homogenity), larger pixel value concentrates on cornerwise degree in representing matrix, it may be assumed that
Wherein: HOM --- gray matrix homogenieity；
Pi,j--- the gray value of some pixel in gray matrix；
I --- gray matrix line number；
J --- gray matrix columns；
B, contrast (Contrast), measurement is varied number local in image, when maximum value in field and minimum value
Between difference increase, its also exponential increase therewith, illustrate image visual effect it is clear whether, it may be assumed that
Wherein: CON --- gray matrix contrast；
Pi,j--- the gray value of some pixel in gray matrix；
I --- gray matrix line number；
J --- gray matrix columns；
C, distinctiveness ratio (Dissmilarity), linearly related to contrast, local contrast is bigger, and diversity is bigger, it may be assumed that
Wherein:
DIS --- gray matrix distinctiveness ratio；
Pi,j--- the gray value of some pixel in gray matrix；
I --- gray matrix line number；
J --- gray matrix columns；
D, entropy (Entroy), entropy are the measurement of information content possessed by image, the size of entropy size and amount of image information
Directly proportional, entropy shows that more greatly the information content of image is abundanter, illustrates that the quality of fusion evaluation is better, it may be assumed that
Wherein: Pi--- all pixel gray value summations of the every a line of gray level co-occurrence matrixes；
E, mean value (Mean) represents the average reflection intensity of all kinds of atural objects in image, is all pixels in remote sensing image
The average value of brightness value, it may be assumed that
Wherein: Pi,j--- the gray value of some pixel in gray matrix；
N --- gray matrix pixel number；
F, standard deviation (Standard Deviation), standard deviation processing is the specific group with reference to adjacent pixel
It closes, it is different with the standard deviation of grey level simple in raw video, it may be assumed that
Wherein:
si,j--- the standard deviation of gray matrix
Pi,j--- the gray value of some pixel in gray matrix；
ui,j--- the gray value of the selected certain picture elements as reference of some in gray matrix；
I --- gray matrix line number；
J --- gray matrix columns；
4, space characteristics
The characteristic element that space characteristics are extracted as auxiliary information, usually used is to collect to obtain in addition to remote sensing image
Electronic land use planning figure, topographic map and geologic map etc..Pass through analysis field of opencast mining and special characteristic atural object element
Between spatial relationship (such as plan range, height difference, contact relation) auxiliary improve field of opencast mining information extraction precision.
Classify to left mixing element, to obtain field of opencast mining information.Random forests algorithm is constructing
When decision tree, using the method for randomly selecting Split Attribute collection, the instruction of 2/3 size is selected from entire training center data set
Practice sample set, 1/3 remaining data estimate the study energy of random forests algorithm as the outer data (Out-of-bag, OOB) of bag
Power, this process are estimated referred to as OOB；Repeat this process ntree time, ntree ∈ [300,500] and be positive integer, finally
Ballot, which is carried out, using the prediction result of ntree decision tree of generation selects final classification results.In addition to this, random forest
The nicety of grading of middle single decision tree also receives the influence of the property set dimension mtry of random sampling, i.e. single is from all features
The characterization factor number and classification randomly selected are concentrated, usual mtry is equal to the evolution of characterization factor sum, Random Forest model
Decision tree number be 300, establish Random Forest model in the present embodiment and participate in operation.
With reference to the regional land use planning chart and digital elevation model, the technological means of use space analysis is eliminated point
Class device logical error, with obtaining four classes such as test section excavating plant (L1), surface plant (L2), Tailings Dam (L3), bare area (L4)
The information extraction of matter environmental element is as a result, carry out precision evaluation simultaneously and using confusion matrix and Kappa coefficient, to information extraction knot
Fruit and test section actual value compare.
It should be noted that Accuracy Assessment provided in an embodiment of the present invention be served only for it is provided in an embodiment of the present invention distant
Feel image terrain classification method precision evaluation, in actual embodiment, due to practical atural object classification data be it is unknown, do not need
It tests to precision.
There are two types of modes at present verifies for nicety of grading: 1. confusion matrix, 2. ROC curve.Confusion matrix is used herein
Method carry out type precision test, the index of evaluation precision includes confusion matrix, overall classification accuracy, Kappa coefficient, every class figure
The cartographic accuracy and user's precision of spot.
By calculating that the field of opencast mining information extracting method provided through the design embodiment obtains as a result, it totally divides
Class precision, Kappa coefficient, user's precision and producer's precision show the field of opencast mining information extracting method of the design embodiment
With very high-precision,
6 remote sensing image ground object information extraction confusion matrix of table
Wherein:
(1) user (user) precision
User's (user) accuracy representing is in the subseries, on classification chart, the check point in the category is fallen in, by just
Really it is classified as the ratio of the category.
(2) drawing (producer) precision
(producer) accuracy representing chart in this subseries, the true reference data in the ground of the category is correctly classified
Probability.
(3) overall accuracy
Overall accuracy refer to the inspection points of all land cover pattern classifications correctly classified it is shared always extract check points
Percentage；Cornerwise all numerical value and the summation divided by whole samples i.e. in confusion matrix.
In formula, PiiThe element of confusion matrix is represented, N represents the summation of classification number.
(4) Kappa coefficient
It is by all pixels for really referring to total (N) multiplied by confusion matrix diagonal line (XKK) sum, then subtract
Really with reference to the product for being classified pixel sum in pixel number and such and then square subtracting respectively divided by pixel sum in all kinds of
In class really with reference to pixel sum and such in be classified the result that the product of pixel sum sums to all categories
In formula, n indicates classification, and N represents the summation (this refers to inspection points) of classification number, XiiIndicate confusion matrix diagonal line
Element, Xi+Indicate classification column summation, X+iIndicate the row summation of classification.
Precision checking computations are carried out by using extracting method of the above method to the present embodiment, it was demonstrated that letter provided in this embodiment
Ceasing extracting method has good precision, can accurately extract the open pit quarry information in satellite image.
It can be applied in practical application.The design embodiment provide based on domestic high resolution ratio satellite remote-sensing image
Field of opencast mining information extracting method by the processing on domestic high resolution ratio satellite remote-sensing image and influences Object hierarchical structure
Building, layer-by-layer exposure mask decomposes removal regional geological environment background information, is based ultimately upon random forest sorting algorithm from mixedly
Species not in extract field of opencast mining information.Final information extraction result precision is higher, has good practicability.
The basic principles, main features and advantages of the present invention have been shown and described above.The technology of the industry
Personnel are it should be appreciated that the present invention is not limited to the above embodiments, and the above embodiments and description only describe this
The principle of invention, various changes and improvements may be made to the invention without departing from the spirit and scope of the present invention, these changes
Change and improvement all fall within the protetion scope of the claimed invention.The claimed scope of the invention by appended claims and its
Equivalent defines.
Claims (10)
1. a kind of field of opencast mining information extracting method based on high-resolution satellite image, it is characterised in that: including following step
It is rapid: 1), to shooting to there is the domestic high-resolution satellite image of target field of opencast mining to pre-process；
2) it, calculates field of opencast mining and appears in the optimum segmentation scale of other regional background environmental elements in satellite image,
Construct the segmentation hierarchical structure of satellite image；
3), based on pretreated satellite image is passed through, calculate the characterization factor collection of each cutting object, the removal of layering exposure mask with
The related regional background environmental element of characterization factor collection；
4) decision tree, is constructed by random forests algorithm, according to characterization factor collection to eliminating area related with characterization factor collection
Remaining mixing element is classified after the background environment element of domain, obtains field of opencast mining information.
2. a kind of field of opencast mining information extracting method based on high-resolution satellite image as described in claim 1, special
Sign is: in the step 1), carrying out pretreated method to satellite image are as follows: have the state of target field of opencast mining to shooting
It produces high-resolution satellite image and carries out the processing operation that radiation calibration, atmospheric correction, ortho-rectification, image co-registration are handled.
3. a kind of field of opencast mining information extracting method based on high-resolution satellite image as described in claim 1, special
Sign is: in the step 2), the method for calculating optimum segmentation scale are as follows: using two index method of Euclid come rational judgment
The optimum segmentation scale of various geological environment elements, for that can not determine the geology ring of differentiation by two index method of Euclid
Border element divides object unit by the method for Image Segmentation.
4. a kind of field of opencast mining information extracting method based on high-resolution satellite image as described in claim 1, special
Sign is: in the step 2), the method that constructs the segmentation hierarchical structure of satellite image are as follows: reference segmentation scale-value is extracted
The principle construction information extraction layer of geological environment element from the easier to the more advanced, by using different segmentations to different geology environmental elements
Scale determines the segmentation scale for being suitble to every kind of geological environment element, establishes reliable classifying rules collection, Wu Faxi on this basis
All geological environment elements divided are set as the bottom.
5. a kind of field of opencast mining information extracting method based on high-resolution satellite image as described in claim 1, special
Sign is: in the step 3), the characterization factor collection includes that spectral signature, geometrical characteristic, textural characteristics and space are special
Sign.
6. a kind of field of opencast mining information extracting method based on high-resolution satellite image as claimed in claim 5, special
Sign is: the spectral signature includes each wave band average gray, each wave band gray standard deviation, brightness value, normalization vegetation
Index and maximum difference.
7. a kind of field of opencast mining information extracting method based on high-resolution satellite image as claimed in claim 5, special
Sign is: the geometrical characteristic includes length-width ratio, density, compact degree, shape index, rectangle fitting degree, asymmetry, circle
Degree.
8. a kind of field of opencast mining information extracting method based on high-resolution satellite image as claimed in claim 5, special
Sign is: the textural characteristics include homogeney, contrast, distinctiveness ratio, entropy, mean value, standard based on gray level co-occurrence matrixes
Difference, similarity and the entropy based on grey scale difference vector, mean value, contrast.
9. a kind of field of opencast mining information extracting method based on high-resolution satellite image as claimed in claim 5, special
Sign is: the space characteristics include away from industrial and mineral warehouse land distance and neighborhood contact relation.
10. a kind of field of opencast mining information extracting method based on high-resolution satellite image as described in claim 1, special
Sign is: in the step 4), passing through the method for random forests algorithm building decision tree are as follows: use and randomly select Split Attribute
The method of collection selects the training sample set of 2/3 size from entire training center data set, and 1/3 remaining data are as number outside bag
It according to come the learning ability of estimating random forests algorithm, repeats this process ntree time, 300≤ntree≤500 finally utilize life
At the prediction result of ntree decision tree carry out ballot and select final classification results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811589599.1A CN109816707A (en) | 2018-12-25 | 2018-12-25 | A kind of field of opencast mining information extracting method based on high-resolution satellite image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811589599.1A CN109816707A (en) | 2018-12-25 | 2018-12-25 | A kind of field of opencast mining information extracting method based on high-resolution satellite image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109816707A true CN109816707A (en) | 2019-05-28 |
Family
ID=66602479
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811589599.1A Pending CN109816707A (en) | 2018-12-25 | 2018-12-25 | A kind of field of opencast mining information extracting method based on high-resolution satellite image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109816707A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110737652A (en) * | 2019-09-30 | 2020-01-31 | 中南大学 | Data cleaning method and system for three-dimensional digital model of surface mine and storage medium |
| CN111178175A (en) * | 2019-12-12 | 2020-05-19 | 中国资源卫星应用中心 | Automatic building information extraction method and system based on high-view satellite image |
| CN111597284A (en) * | 2020-04-29 | 2020-08-28 | 武汉子雨科技有限公司 | Method for rapidly extracting surveying and mapping geographic information data |
| CN111723711A (en) * | 2020-06-10 | 2020-09-29 | 内蒙古农业大学 | Plianes and object-oriented mulching film information extraction method and system |
| CN112329706A (en) * | 2020-11-23 | 2021-02-05 | 京师天启(北京)科技有限公司 | Mining land identification method based on remote sensing technology |
| CN113642579A (en) * | 2021-07-08 | 2021-11-12 | 湖南元想科技有限公司 | Method and system for determining and dividing triarrhena sacchariflora growing region |
| CN114549972A (en) * | 2022-01-17 | 2022-05-27 | 中国矿业大学(北京) | Strip mine stope extraction method, apparatus, device, medium, and program product |
| CN116363502A (en) * | 2023-01-31 | 2023-06-30 | 中国科学院地理科学与资源研究所 | Mining land multidimensional information acquisition method and device integrating multisource geographic big data |
| CN116503677A (en) * | 2023-06-28 | 2023-07-28 | 武汉大学 | Wetland classification information extraction method, system, electronic equipment and storage medium |
| CN117520965A (en) * | 2024-01-04 | 2024-02-06 | 华洋通信科技股份有限公司 | Industrial and mining operation data classification method based on artificial intelligence |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105608473A (en) * | 2015-12-31 | 2016-05-25 | 中国资源卫星应用中心 | High-precision land cover classification method based on high-resolution satellite image |
| CN105608474A (en) * | 2015-12-30 | 2016-05-25 | 二十一世纪空间技术应用股份有限公司 | High-resolution-image-based regional adaptive cultivated land extraction method |
| CN108492288A (en) * | 2018-03-15 | 2018-09-04 | 武汉大学 | The high score satellite image change detecting method of multiple dimensioned stratified sampling based on random forest |
-
2018
- 2018-12-25 CN CN201811589599.1A patent/CN109816707A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105608474A (en) * | 2015-12-30 | 2016-05-25 | 二十一世纪空间技术应用股份有限公司 | High-resolution-image-based regional adaptive cultivated land extraction method |
| CN105608473A (en) * | 2015-12-31 | 2016-05-25 | 中国资源卫星应用中心 | High-precision land cover classification method based on high-resolution satellite image |
| CN108492288A (en) * | 2018-03-15 | 2018-09-04 | 武汉大学 | The high score satellite image change detecting method of multiple dimensioned stratified sampling based on random forest |
Non-Patent Citations (1)
| Title |
|---|
| 张凯翔: "露天矿地质环境解译标志体系和信息提取方法的研究及示范应用", 《中国博士学位论文全文数据库 基础科学辑》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110737652B (en) * | 2019-09-30 | 2022-03-11 | 中南大学 | Data cleaning method and system for three-dimensional digital model of surface mine and storage medium |
| CN110737652A (en) * | 2019-09-30 | 2020-01-31 | 中南大学 | Data cleaning method and system for three-dimensional digital model of surface mine and storage medium |
| CN111178175A (en) * | 2019-12-12 | 2020-05-19 | 中国资源卫星应用中心 | Automatic building information extraction method and system based on high-view satellite image |
| CN111597284A (en) * | 2020-04-29 | 2020-08-28 | 武汉子雨科技有限公司 | Method for rapidly extracting surveying and mapping geographic information data |
| CN111723711A (en) * | 2020-06-10 | 2020-09-29 | 内蒙古农业大学 | Plianes and object-oriented mulching film information extraction method and system |
| CN112329706A (en) * | 2020-11-23 | 2021-02-05 | 京师天启(北京)科技有限公司 | Mining land identification method based on remote sensing technology |
| CN113642579A (en) * | 2021-07-08 | 2021-11-12 | 湖南元想科技有限公司 | Method and system for determining and dividing triarrhena sacchariflora growing region |
| CN114549972A (en) * | 2022-01-17 | 2022-05-27 | 中国矿业大学(北京) | Strip mine stope extraction method, apparatus, device, medium, and program product |
| CN114549972B (en) * | 2022-01-17 | 2023-01-03 | 中国矿业大学(北京) | Strip mine stope extraction method, device, equipment and medium |
| CN116363502A (en) * | 2023-01-31 | 2023-06-30 | 中国科学院地理科学与资源研究所 | Mining land multidimensional information acquisition method and device integrating multisource geographic big data |
| CN116363502B (en) * | 2023-01-31 | 2023-10-20 | 中国科学院地理科学与资源研究所 | Mining land multidimensional information acquisition method and device integrating multisource geographic big data |
| CN116503677A (en) * | 2023-06-28 | 2023-07-28 | 武汉大学 | Wetland classification information extraction method, system, electronic equipment and storage medium |
| CN116503677B (en) * | 2023-06-28 | 2023-09-05 | 武汉大学 | Wetland classification information extraction method, system, electronic equipment and storage medium |
| CN117520965A (en) * | 2024-01-04 | 2024-02-06 | 华洋通信科技股份有限公司 | Industrial and mining operation data classification method based on artificial intelligence |
| CN117520965B (en) * | 2024-01-04 | 2024-04-09 | 华洋通信科技股份有限公司 | Industrial and mining operation data classification method based on artificial intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109816707A (en) | A kind of field of opencast mining information extracting method based on high-resolution satellite image | |
| CN101840581B (en) | Method for extracting profile of building from satellite remote sensing image | |
| Saha et al. | An object-oriented approach to automated landform mapping: A case study of drumlins | |
| Erenoglu et al. | An UAS-assisted multi-sensor approach for 3D modeling and reconstruction of cultural heritage site | |
| Pesaresi et al. | Toward global automatic built-up area recognition using optical VHR imagery | |
| Adam et al. | A comparison of pixel-based and object-based approaches for land use land cover classification in semi-arid areas, Sudan | |
| Milani et al. | Robust quantification of riverine land cover dynamics by high-resolution remote sensing | |
| CN106846322B (en) | The SAR image segmentation method learnt based on curve wave filter and convolutional coding structure | |
| Xu et al. | Delineation of cultivated land parcels based on deep convolutional networks and geographical thematic scene division of remotely sensed images | |
| Yan | Pixel based and object oriented image analysis for coal fire research | |
| CN111339827A (en) | SAR image change detection method based on multi-region convolutional neural network | |
| Chen et al. | Detecting changes in high-resolution satellite coastal imagery using an image object detection approach | |
| Xu et al. | Feature-based constraint deep CNN method for mapping rainfall-induced landslides in remote regions with mountainous terrain: An application to Brazil | |
| Salati et al. | Lithological mapping and fuzzy set theory: Automated extraction of lithological boundary from ASTER imagery by template matching and spatial accuracy assessment | |
| Li et al. | An improved approach of information extraction for earthquake-damaged buildings using high-resolution imagery | |
| Bengoufa et al. | Rocky shoreline extraction using a deep learning model and object-based image analysis | |
| CN109829426A (en) | Railway construction temporary building monitoring method and system based on high score remote sensing image | |
| Sugg et al. | Mapping impervious surfaces using object-oriented classification in a semiarid urban region | |
| Devi et al. | Change detection techniques–A survey | |
| Liu et al. | Extraction of earthquake-induced collapsed buildings from bi-temporal VHR images using object-level homogeneity index and histogram | |
| An et al. | Object-oriented urban dynamic monitoring—A case study of Haidian District of Beijing | |
| Tercan et al. | Effectiveness of autoencoder for lake area extraction from high-resolution RGB imagery: an experimental study | |
| Zhang et al. | Application and evaluation of object-oriented technology in high-resolution remote sensing image classification | |
| Gonçalves et al. | Assessment of the state of conservation of buildings through roof mapping using very high spatial resolution images | |
| Yamashkin et al. | Classification of Metageosystems by Ensembles of Machine Learning Models |
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 |
Application publication date: 20190528 |
|
| RJ01 | Rejection of invention patent application after publication |