CN107392926A - Characteristics of remote sensing image system of selection based on soil thematic map early stage - Google Patents
Characteristics of remote sensing image system of selection based on soil thematic map early stage Download PDFInfo
- Publication number
- CN107392926A CN107392926A CN201710839939.0A CN201710839939A CN107392926A CN 107392926 A CN107392926 A CN 107392926A CN 201710839939 A CN201710839939 A CN 201710839939A CN 107392926 A CN107392926 A CN 107392926A
- Authority
- CN
- China
- Prior art keywords
- feature
- remote sensing
- soil
- image
- early stage
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
- G06T2207/20152—Watershed segmentation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of characteristics of remote sensing image system of selection based on soil thematic map early stage, comprise the following steps：A kind of image feature set of Remote Sensing Image Segmentation algorithm and cutting object is selected, and corresponding partitioning parameters and feature extraction algorithm are set；Soil thematic map early stage is cut into the figure layer of respective amount by plot classification, and extracts influence intensity map and feature weight matrix, generates feature weight distribution map；Remote sensing image to be analyzed is split, extracts the image feature of cutting object；A kind of image classification device is selected, and the weight for improving its characteristic of division is set, the grader of construction feature weighting；It is the object setting feature weight in segmentation figure according to feature weight distribution map, and its image feature and weight is input in weighted feature classification device, calculates the land status of cutting object, generate soil thematic map.The present invention can realize the adaptive optimization adjustment of remote sensing image local features weight, can improve remote sensing images analysis precision.
Description
Technical field
The present invention relates to a kind of characteristics of remote sensing image system of selection based on soil thematic map early stage, belong to object-oriented
Remote sensing images analysis field.
Background technology
With the fast development of high (space) resolution remote sense, the remote sensing analysis (Object-Based of object-oriented
Image Analysis, OBIA) method turn into current remote sensing application technical way.Relative to traditional based on pixel
(pixel-based) remote sensing analysis technology, the image analysis methods of object-oriented on the basis of using image spectral signature,
High resolution image abundant geometric properties, textural characteristics and Features In Pattern of Spatial etc. can be fully excavated, and can further be incorporated
The high level knowledge such as social economy, spatial model, realizes the image analysing computer of higher precision and higher efficiency.Corresponding object-oriented
The bibliography of remote sensing analysis includes, Zhou Yanan, Luo Jiancheng, and the adaptive remote sensing image that the such as Cheng Xi multiple features incorporate is multiple dimensioned
Split [J] Wuhan University Journals：Information science version, 2013,38 (1):The high-resolution satellites such as 19-22, Zhou Chenghu, Luo Jiancheng
Remote sensing image geoscience computing [M] Beijing:Science Press, 2009, Blaschke, T., 2010.Object based image
analysis for remote sensing.ISPRS Journal of Photogrammetry and Remote
Sensing,65(1):2-16、Myint,S.W.,Gober,P.,Brazel,A.,Grossman-Clarke,S.,Weng,Q.,
2011.Per-pixel vs.object-based classification of urban land cover extraction
using high spatial resolution imagery.Remote Sens.Environ.115(5):1145-1161 etc..
But in the remote sensing analysis of object-oriented, various image features (including spectral signature, geometric properties,
Textural characteristics and spatial relationship etc.) it is constructed and extracts, new challenge is proposed to the analysis method of object-oriented.One side
Face, extraction and the image feature using more cutting objects can consume more processing and analysis time；On the other hand, relatively
In a limited number of training samples, excessive image feature can reduce the precision of image analysing computer on the contrary.Therefore, researcher attempts to set
The method that meter reduces characteristic dimension, to find most important feature in feature space, to be obtained using less characteristics of objects
Obtain the analysis precision suitable with utilizing whole features.It is (special in Digital Image Processing and analysis field, common reduction characteristic dimension
Sign dimensionality reduction) method can be divided into feature extraction and the major class of feature selecting two.In feature extraction, original characteristics of objects space quilt
The feature space of new relatively low dimensional is converted into, to build and extract prior feature.Yet with feature extracting method
Original feature space is changed, therefore, it is difficult to provide the theoretical explanation to dimensionality reduction result；Therefore more usually feature selecting
Method (being also exactly present invention method of interest).The target of feature selecting is to choose lesser amt from original characteristic set
, prior feature composition character subset, and this feature subset is approached the analysis of even more than original characteristic set
Precision.It is to utilize the attribute information between feature either based on the evaluation to classification results according to method, the spy in remote sensing application
Sign system of selection is roughly divided into three classes：Filtration method (filter), pack (wrapper) and embedding inlay technique (embedded).Filtering
Statistical property (such as correlation, comentropy) between method feature based carries out importance ranking to feature, and selects importance
Higher combinations of features, without evaluating the classification results in later stage.Therefore the efficiency of filtration method is very high but is difficult to directly to dividing
Class result optimizes.Correspondingly, parcel rule is assessed by selecting at random using some machine learning methods (such as SVM classifier)
The quality of the classification results of the character subset generation taken, and attempt to find the combinations of features of generation optimal result.Although such side
Method can obtain preferably analysis prediction effect, but it needs to attempt many character subsets and less efficient.Being embedded in rule is
To filtration method and the compromise of pack, it just completes the evaluation and selection of feature while construction feature preference pattern, therefore
Without the subsequently evaluation to selected combinations of features.Such as random forest (random forest) algorithm is exactly a kind of conventional embedding
Enter method, it is by calculating feature to the average influence degree of classification results come the importance of evaluating characteristic.Corresponding Feature Dimension Reduction
Bibliography include, Zhong P, Zhang P, Wang R.Dynamic learning of SMLR for feature
selection and classification of hyperspectral data[J].IEEE Geoscience and
Remote Sensing Letters,2008,5(2):280-284、Donoho D L.High-dimensional data
analysis:The curses and blessings of dimensionality[J].AMS Math Challenges
Lecture,2000,1:32、Guyon I,Elisseeff A.An Introduction to Variable Feature
Selection[J].Journal of Machine Learning Research,2003,3:1157-1182、Pal M,
Foody G M.Feature selection for classification of hyperspectral data by SVM
[J].IEEE Transactions on Geoscience and Remote Sensing,2010,48(5):2297-2307、
Jain A,Zongker D.Feature selection:Evaluation,application,and small sample
performance[J].IEEE transactions on pattern analysis and machine
intelligence,1997,19(2):153-158、Duro D C,Franklin S E,DubéM G.Multi-scale
object-based image analysis and feature selection of multi-sensor earth
observation imagery using random forests[J].International Journal of Remote
Sensing,2012,33(14):4502-4526、Ma L,Fu T,Blaschke T,et al.Evaluation of
Feature Selection Methods for Object-Based Land Cover Mapping of Unmanned
Aerial Vehicle Imagery Using Random Forest and Support Vector Machine
Classifiers[J].ISPRS International Journal of Geo-Information,2017,6(2):51。
Although scholar proposes substantial amounts of feature selection approach, numerous feature selecting experiments is also carried out；It is but most of
Method and experiment be all non-supervisory (without the support of priori), it is globally consistent (towards the optimal of whole test block
Feature selecting is difficult to do adaptive optimization for the local characteristic of image), therefore the lifting to image analysing computer effect is limited.
In summary, in current visible patent and literature research, most feature selection approach is all non-supervisory
, it is global fixed, and can effectively incorporate priori and can local auto-adaptive optimization feature selection approach it is less, especially
It is to be answered in the involvement mechanism of priori, the adaptive optimization method of feature weight and flow, feature weight remote sensing analysis
With the methods of remain more the problem of, more do not develop practicable solution.
The content of the invention
The technical problem to be solved by the invention is to provide it is a kind of can improve remote sensing images analysis precision based on early stage
The characteristics of remote sensing image system of selection of soil thematic map.
In order to solve the above technical problems, the technical solution adopted by the present invention is：
Based on the characteristics of remote sensing image system of selection of soil thematic map early stage, comprise the following steps：
(1) a kind of image feature set of Remote Sensing Image Segmentation algorithm and cutting object is selected, and corresponding segmentation is set
Parameter and feature extraction algorithm；
(2) soil thematic map early stage is cut into the figure layer of respective amount by plot classification, and extracts every kind of land status
Influence intensity map；
(3) soil thematic map early stage is registrated on its corresponding remote sensing image early stage, according to the feature described in step (1)
The image feature in plot in extraction algorithm extraction soil thematic map early stage, and store into the attribute list of thematic map, by plot class
Soil thematic map early stage is not cut into the figure layer of respective amount, and extracts the feature weight matrix of every kind of land status；
(4) spy of the influence intensity map in step (2) and the feature weight matrix generation test block in step (3) is combined
Levy weight distribution figure；
(5) partitioning algorithm according to step (1) and feature extraction algorithm divide remote sensing image to be analyzed
Cut, extract the image feature of cutting object, and store into the attribute list of segmentation figure；
(6) a kind of image classification device is selected, and the weight for improving its characteristic of division is set, the classification of construction feature weighting
Device；
(7) be that object in segmentation figure sets feature weight based on feature weight distribution map described in step (4), and by its
Image feature and weight are input in weighted feature classification device, calculate the land status of cutting object, generate soil thematic map.
Partitioning algorithm described in step (1) includes fractional spins, mean shift segmentation algorithm or multiresolution point
Cut algorithm.
Image feature set described in step (1) includes spectral signature, geometric properties and textural characteristics.
Trifling region rejecting is carried out to single figure layer in step (2) and vector to raster conversion generates bianry image, then to two
Value image carries out range conversion and then generates the influence intensity map of this kind of land status, other figure layers is equally handled, i.e.,
Obtain the influence intensity map of each land status.
The two-value reclassification of " target-background " is carried out in step (3) to single figure layer, with the atural object generic with figure layer
For target, other atural objects are the class figure of Background generation two, then calculate target class another characteristic power using random forests algorithm
Other figure layers are equally handled, that is, obtain feature weight matrix by weight vector.
Image classification device described in step (6) includes minimum distance classifier, Bayes classifier or k nearest neighbor grader.
The beneficial effect that the present invention is reached：
Soil thematic map (such as land use/cover figure, the vegetation thematic map etc.) record of early stage has atural object in test block
Various attribute (information such as feature, classification, space size and distribution)；And in the short time regional area land use/cover
Matter also change it is smaller, therefore early stage soil thematic map can as a kind of priori be used for remote sensing image analysis and understanding.Will
Early stage, soil thematic map was registrated on remote sensing image to be analyzed, and to carry out assisted Selection more excellent for the priori for extracting thematic map
Combinations of features (for example assign higher feature weight in water body in lake area, the spectral signature and textural characteristics of atural object；And in city
Area, the shape facility of atural object assign higher weight).On the other hand, geographical space scene is complicated, its by different scale,
Different classes of, different attribute atural object combines；And the atural object property of regional area is different from the property in integral experiment area, no
Atural object property with regional area is also not quite similar.Therefore suitable for integral experiment area feature choice subsets not necessarily certain
The optimal characteristics combination of individual regional area；Need to carry out image regional area more targeted and Objective feature choosing
Select, further improve feature selecting and the effect of image analysing computer.The present invention is based on soil thematic map early stage and its corresponding early stage
Remote sensing image, can effectively extract the influence intensity map and feature weight matrix of each land status, and construct the feature of trial zone
Weight map；And then the grader of binding characteristic weight, the adaptive optimization of feature weight in classification is realized, improves remote sensing image point
The precision of analysis.Adaptive features select method involved in the present invention is applicable not only to the classification of remote sensing image, is also suitable for it
His remote sensing images analysis, the segmentation of such as remote sensing image, the thematic map extraction of remote sensing image and target identification.
Brief description of the drawings
Fig. 1 is the schematic diagram of the characteristics of remote sensing image selection under soil thematic map support early stage；
Fig. 2 is the implementation process figure of the characteristics of remote sensing image selection under soil thematic map support early stage；
Fig. 3 is the flow chart of Remote Sensing Image Segmentation (by taking average drifting as an example) and cutting object feature extraction；
Fig. 4 is the flow chart of distance (influence) intensity map generation of (each) land status；
Fig. 5 is the flow chart that (each) land status feature weight calculates；
Fig. 6 is the schematic diagram of feature weight figure generation；
Fig. 7 is the land classification effect diagram of somewhere remote sensing image.
Embodiment
The invention will be further described below in conjunction with the accompanying drawings.Following examples are only used for clearly illustrating the present invention
Technical scheme, and can not be limited the scope of the invention with this.
Fig. 1 is the schematic diagram of the characteristics of remote sensing image selection under soil thematic map support early stage, wherein calculating feature power
In the step of multigraph, extracted from soil thematic map early stage and its corresponding remote sensing image each land status apart from intensity map and soil
The feature weight matrix of ground classification, and build the feature weight figure of test block；In the classifying step of characteristic weighing, according to segmentation
The locus of object to be analyzed in figure, inquires about the weight of its each feature from feature weight figure, and sets grader with this
Weight；Realize that the self-adaptive features of remote sensing image are preferred with this；
Fig. 2 is the particular flow sheet of the characteristics of remote sensing image selection under soil thematic map support early stage, includes remote sensing shadow
The segmentation of picture and feature extraction, distance (influence) the intensity map generalization of each land status, the feature weight matrix of land status
Structure, the calculating of feature weight (distribution) figure, the design of weighted feature classification device, soil thematic classification map generalization etc. 6
Realize unit；
As shown in figure 3, in Remote Sensing Image Segmentation (specific implementation of the invention is by taking mean shift segmentation algorithm as an example) and spy
In sign extraction.Image Segmentation based on average drifting generally comprises two steps of images filter and region merging technique；Wherein average is floated
Move algorithm and apply to filtering, it is therefore intended that find the Local Extremum in image and generate the imagery zone of average；Merge
Process is then to find the connected region of average block and form final cutting object.In filtering, joint considers image
Locus and spectral signature, form joint vector x=(x of a d=p+2 dimensions,xr), wherein xsRepresent image in grid/
The position coordinates of pixel, xrRepresent the p dimensional vector features of grid/pixel in image.Mean shift algorithm is general according to kernel function
Rate density estimation, Local Extremum is found in above-mentioned vector space.Wherein multidimensional kernel function is defined as：
Wherein k (x) is the spatially and spectrally kernel function in domain, hs、hrThe nucleus band of respectively spatial domain and spectral domain is wide, and C is
Generalized constant.So mean shift iterations function is then：
Wherein xtFor by the position of t iteration rear mold point (mode), wiFor the effect weight of pixel i in x fields.Repeatedly
The track that the positions passed through of x in generation, i.e. sequence { x, m (x), m (m (x)) ... } are x；Average drifting is always pointed at most
The place of big local density, drift value goes to zero at density function maximum, then iteration terminates.Mean filter is special by calculating
The weighted sum of local (in neighborhood) sample point is drifted about to realize in sign space；On the one hand the pixel inside atural object in image is carried out
Smoothing processing, the boundary characteristic of atural object is on the other hand remained again.First with transitive closure in region merging technique
(transitive closure) algorithm iteration by spatially adjacent two it is (multiple) have parallel pattern (mode) and
The homogenous area of smaller boundary intensity (edge strength) merges, to generate bigger cut zone.Then by less point
Cut region (number of pixels is less than S) to be merged into adjacent cut zone, generate final cutting object.
On the basis of above-mentioned mean shift segmentation, the spectral signature of cutting object, textural characteristics, shape facility etc. are extracted
A variety of attributive character, it is specific as shown in table 1.
The image feature of the cutting object of table 1.
As shown in figure 4, in distance (influence) intensity map of generation (each) land status.Range conversion is used to describe two-value
Backdrop pels are then have recorded in bianry image to the most narrow spacing of region of interest with the separation degree of region of interest, range conversion figure in image
From.The present invention, for region of interest, (the water removal of non-region of interest is calculated using range conversion with the soil region (such as water body) of certain classification
External atural object is such as ploughed, forest land) minimum range of region of interest is arrived, and the influence (distance) for being used to portray the land type is strong
Degree figure；And then generate the influence intensity map of (each) land status.Soil thematic map is cut in the specific implementation, being first according to classification
It is divided into multiple figure layers (a kind of corresponding figure layer of land status)；Then trifling region rejecting and vector grid are carried out to single figure layer
Format and generate bianry image；Then influence (distance) intensity that range conversion generates this kind of land status is carried out to bianry image
Figure；Finally other land status are carried out with same processing, that is, obtains the influence intensity map of each land status in test block.To reality
The object S on any locus in area is tested, corresponding classification intensity can be found from the influence intensity map of land status
Value, and the classification that linear normalization is obtained at the point influences intensity vector D=(di, i=1,2 ..., N, wherein 0<di<1,∑di
=1, N are the number of land status).
As shown in figure 5, in the feature weight for calculating land status.Random forest is a kind of machine for integrating more decision trees
Device learning method, and the importance returned with variable (feature) in classification problem can be assessed.The present invention will utilize random forest
Algorithm calculates the significance level of the various features of each land type (weight).In the specific implementation, first by the soil of early stage
Thematic map geometric maps are to a variety of image features (such as table 1 on the satellite image of early stage, calculating each plot in soil thematic map
It is shown)；Then it is multiple figure layers (a kind of corresponding figure layer of land status) by soil thematic map cutting according to classification；Then it is right
Single figure layer carries out the two-value reclassification (and the generic atural object of figure layer is target, and other atural objects are background) of " target-background "
Two class figures are generated, and the target class another characteristic weight vectors are estimated using random forests algorithm；Finally other figure layers are carried out
Same processing, that is, obtain feature weight matrix W=(w of land statusij, i=1,2 ..., N, j=1,2 ..., M, wherein 0<
wij<1, N is the number of land status, and M is the number of image feature).
As shown in fig. 6, in feature weight (distribution) figure is generated.According to flow shown in Fig. 4 can experiment with computing area it is any
The influence intensity map of the various land types of opening position, the feature weight matrix of land type can be calculated according to flow shown in Fig. 5.
The feature weight vector WF=(wf at the S of optional position can be further calculated using formula (1)j, j=1,2 ..., M, wherein M
It is the number of image feature), wherein diFor the image intensity of i-th kind of land type, wijFor the jth Wei Te of i-th kind of land type
The weight of sign.
Grader (specific implementation of the invention is by taking k nearest neighbor sorting algorithm as an example) and feature extraction in construction feature weighting
In) in.K nearest neighbor (K-nearest neighbor, K-NN) grader only considers the spatial distribution of priori sample and ignored to class
Other probability distribution it is assumed that being one of conventional Remote Image Classification.In K-NN assorting processes, test is searched first
K closest priori sample of sample, the rule for being then based on majority voting scheme judge according to the classification of priori sample
The classification of test sample.The present invention carrys out the grader of construction feature weighting by K-NN classification is transformed.In the specific implementation, utilizing
Formula (2) calculates test sample to the nearest neighbor distance of priori sample, wherein Δ fjFor the characteristic distance of jth dimension.
In standard K-NN classification, wfjIt is all set to 1.0 (waiting power to set)；And in the improved K-NN classification of characteristic weighing
, it is necessary to set the weight wf of sample different characteristic in devicej；Realize that the effect of the key character of weighted feature classification device is big, non-
Key character acts on small design it is assumed that providing method support to be subsequently generated soil thematic map.
In soil thematic classification figure is generated.According to the object-oriented Classification in Remote Sensing Image process of routine, test block is carried out successively
Training sample and checking the selecting of sample, the K-NN of improved weighted feature classification, the checking of classification results etc..Improved
, it is necessary to inquire the weight of each image feature from feature weight figure according to the locus of sample to be sorted in K-NN classification,
And it is brought into the kind judging for improving K-NN.
Fig. 7 illustrates the soil thematic classification effect for the somewhere satellite image implemented using the method in the present invention.From whole
From the point of view of body classifying quality, the simple region such as water body of the inventive method either in image, forest land is still used in the construction of image
The complex regions such as ground, irrigated land, road can obtain preferable classifying quality.
The example of the present invention is realized on a pc platform.It the experiment proved that, the present invention can effectively assess test block
The feature weight in middle local image region, and in follow-up object oriented classification, more conventional global characteristics system of selection (ratio
Such as side's of card selection (Chi-square) and recursion elimination (Recursive Feature Elimination)) have largely
Precision improve that (present invention give the land classification precision of example image, relative to card side's system of selection raising about 8%, relatively
Improved about 9%) in recursion elimination method.Method mentioned by the present invention can be widely applied to high-resolution remote sensing image towards
During image analysing computer, classification and identification of object etc., such as Third National agricultural census, survey of territorial resources are large-scale should
With.
Described above is only the preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, without departing from the technical principles of the invention, some improvement and deformation can also be made, these are improved and deformation
Also it should be regarded as protection scope of the present invention.
Claims (6)
1. based on the characteristics of remote sensing image system of selection of soil thematic map early stage, it is characterized in that, comprise the following steps：
（1）A kind of image feature set of Remote Sensing Image Segmentation algorithm and cutting object is selected, and corresponding partitioning parameters are set
And feature extraction algorithm；
（2）Soil thematic map early stage is cut into the figure layer of respective amount by plot classification, and extracts the shadow of every kind of land status
Ring intensity map；
（3）Soil thematic map early stage is registrated on its corresponding remote sensing image early stage, according to step（1）Described feature extraction
The image feature in plot in algorithm extraction soil thematic map early stage, and store into the attribute list of thematic map, will by plot classification
Early stage, soil thematic map was cut into the figure layer of respective amount, and extracted the feature weight matrix of every kind of land status；
（4）With reference to step（2）In influence intensity map and step（3）In feature weight matrix generation test block feature power
Redistribution figure；
（5）According to step（1）Described partitioning algorithm and feature extraction algorithm is split to remote sensing image to be analyzed, carries
The image feature of cutting object is taken, and is stored into the attribute list of segmentation figure；
（6）A kind of image classification device is selected, and the weight for improving its characteristic of division is set, the grader of construction feature weighting；
（7）Based on step（4）Described in feature weight distribution map be that object in segmentation figure sets feature weight, and by its image
Feature and weight are input in weighted feature classification device, calculate the land status of cutting object, generate soil thematic map.
2. the characteristics of remote sensing image system of selection according to claim 1 based on soil thematic map early stage, it is characterized in that, step
Suddenly（1）Described in partitioning algorithm include fractional spins, mean shift segmentation algorithm or multi-resolution segmentation algorithm.
3. the characteristics of remote sensing image system of selection according to claim 1 based on soil thematic map early stage, it is characterized in that, step
Suddenly（1）Described in image feature set include spectral signature, geometric properties and textural characteristics.
4. the characteristics of remote sensing image system of selection according to claim 1 based on soil thematic map early stage, it is characterized in that, step
Suddenly（2）In trifling region rejecting is carried out to single figure layer and vector to raster conversion generates bianry image, then bianry image is carried out
Other figure layers are equally handled, that is, obtain each soil by range conversion and then the influence intensity map for generating this kind of land status
The influence intensity map of classification.
5. the characteristics of remote sensing image system of selection according to claim 1 based on soil thematic map early stage, it is characterized in that, step
Suddenly（3）In to single figure layer carry out " target-background " two-value reclassification, using the atural object generic with figure layer as target, other
Atural object is the class figure of Background generation two, then the target class another characteristic weight vectors is calculated using random forests algorithm, to it
He is equally handled figure layer, that is, obtains feature weight matrix.
6. the characteristics of remote sensing image system of selection according to claim 1 based on soil thematic map early stage, it is characterized in that, step
Suddenly（6）Described in image classification device include minimum distance classifier, Bayes classifier or k nearest neighbor grader.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710839939.0A CN107392926B (en) | 2017-09-18 | 2017-09-18 | Remote sensing image feature selection method based on early-stage land thematic map |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710839939.0A CN107392926B (en) | 2017-09-18 | 2017-09-18 | Remote sensing image feature selection method based on early-stage land thematic map |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107392926A true CN107392926A (en) | 2017-11-24 |
| CN107392926B CN107392926B (en) | 2020-12-01 |
Family
ID=60350563
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710839939.0A Active CN107392926B (en) | 2017-09-18 | 2017-09-18 | Remote sensing image feature selection method based on early-stage land thematic map |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107392926B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108376232A (en) * | 2018-01-04 | 2018-08-07 | 北京星衡科技有限公司 | A kind of method and apparatus of automatic interpretation for remote sensing image |
| CN109657616A (en) * | 2018-12-19 | 2019-04-19 | 四川立维空间信息技术有限公司 | A kind of remote sensing image land cover pattern automatic classification method |
| 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 |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6363161B2 (en) * | 1998-10-23 | 2002-03-26 | Facet Technology Corp. | System for automatically generating database of objects of interest by analysis of images recorded by moving vehicle |
| CN101661497A (en) * | 2009-09-10 | 2010-03-03 | 北京农业信息技术研究中心 | Remote sensing land use change detection method and system thereof |
| CN101840581A (en) * | 2010-02-05 | 2010-09-22 | 北京交通大学 | Method for extracting profile of building from satellite remote sensing image |
| CN101930547A (en) * | 2010-06-24 | 2010-12-29 | 北京师范大学 | Method for automatically classifying remote sensing image based on object-oriented unsupervised classification |
| CN102024258A (en) * | 2010-12-15 | 2011-04-20 | 中国科学院遥感应用研究所 | Multi-scale segmentation method for remote sensing image with boundary maintenance characteristics |
| CN103839261A (en) * | 2014-02-18 | 2014-06-04 | 西安电子科技大学 | SAR image segmentation method based on decomposition evolution multi-objective optimization and FCM |
-
2017
- 2017-09-18 CN CN201710839939.0A patent/CN107392926B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6363161B2 (en) * | 1998-10-23 | 2002-03-26 | Facet Technology Corp. | System for automatically generating database of objects of interest by analysis of images recorded by moving vehicle |
| CN101661497A (en) * | 2009-09-10 | 2010-03-03 | 北京农业信息技术研究中心 | Remote sensing land use change detection method and system thereof |
| CN101840581A (en) * | 2010-02-05 | 2010-09-22 | 北京交通大学 | Method for extracting profile of building from satellite remote sensing image |
| CN101930547A (en) * | 2010-06-24 | 2010-12-29 | 北京师范大学 | Method for automatically classifying remote sensing image based on object-oriented unsupervised classification |
| CN102024258A (en) * | 2010-12-15 | 2011-04-20 | 中国科学院遥感应用研究所 | Multi-scale segmentation method for remote sensing image with boundary maintenance characteristics |
| CN103839261A (en) * | 2014-02-18 | 2014-06-04 | 西安电子科技大学 | SAR image segmentation method based on decomposition evolution multi-objective optimization and FCM |
Non-Patent Citations (2)
| Title |
|---|
| 郝虑远: ""基于纹理分析的高分辨率影像面向对象分类研究"", 《中国优秀硕士学位论文全文数据库（电子期刊）基础科学辑》 * |
| 陈建容: ""面向对象的土地利用/覆盖信息提取"", 《中国优秀硕士学位论文全文数据库（电子期刊）基础科学辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108376232A (en) * | 2018-01-04 | 2018-08-07 | 北京星衡科技有限公司 | A kind of method and apparatus of automatic interpretation for remote sensing image |
| CN109657616A (en) * | 2018-12-19 | 2019-04-19 | 四川立维空间信息技术有限公司 | A kind of remote sensing image land cover pattern automatic classification method |
| 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107392926B (en) | 2020-12-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wu et al. | ORSIm detector: A novel object detection framework in optical remote sensing imagery using spatial-frequency channel features | |
| CN108038445B (en) | SAR automatic target identification method based on multi-view deep learning framework | |
| Fu et al. | Using convolutional neural network to identify irregular segmentation objects from very high-resolution remote sensing imagery | |
| Chen et al. | Vehicle detection in high-resolution aerial images based on fast sparse representation classification and multiorder feature | |
| Wang et al. | Land cover change detection at subpixel resolution with a Hopfield neural network | |
| Lu et al. | A survey of image classification methods and techniques for improving classification performance | |
| Liu et al. | Unsupervised change detection in multispectral remote sensing images via spectral-spatial band expansion | |
| CN106897679A (en) | A kind of semantic change detecting method and system based on improvement fuzzy C-means clustering | |
| Asokan et al. | Machine learning based image processing techniques for satellite image analysis-a survey | |
| CN107067405B (en) | Remote sensing image segmentation method based on scale optimization | |
| Liu et al. | Enhancing spectral unmixing by local neighborhood weights | |
| CN107784657A (en) | A kind of unmanned aerial vehicle remote sensing image partition method based on color space classification | |
| Li et al. | A review on deep learning techniques for cloud detection methodologies and challenges | |
| Zhang et al. | Fusion of heterogeneous earth observation data for the classification of local climate zones | |
| CN105976376A (en) | High resolution SAR image target detection method based on part model | |
| Jiang et al. | Intelligent image semantic segmentation: a review through deep learning techniques for remote sensing image analysis | |
| Han et al. | The edge-preservation multi-classifier relearning framework for the classification of high-resolution remotely sensed imagery | |
| CN107392926A (en) | Characteristics of remote sensing image system of selection based on soil thematic map early stage | |
| Chen et al. | Landslide detection using probability regression, a case study of Wenchuan, northwest of Chengdu | |
| CN107463944B (en) | A kind of road information extracting method using multidate High Resolution SAR Images | |
| Alhichri et al. | Tile‐Based Semisupervised Classification of Large‐Scale VHR Remote Sensing Images | |
| Khesali et al. | Semi automatic road extraction by fusion of high resolution optical and radar images | |
| Sutha | Object based classification of high resolution remote sensing image using HRSVM-CNN classifier | |
| Al-Ghrairi et al. | Classification of satellite images based on color features using remote sensing | |
| Kazimi et al. | Semantic segmentation of manmade landscape structures in digital terrain 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |