CN105608474B - Region adaptivity plant extraction method based on high resolution image - Google Patents
Region adaptivity plant extraction method based on high resolution image Download PDFInfo
- Publication number
- CN105608474B CN105608474B CN201511034394.3A CN201511034394A CN105608474B CN 105608474 B CN105608474 B CN 105608474B CN 201511034394 A CN201511034394 A CN 201511034394A CN 105608474 B CN105608474 B CN 105608474B
- Authority
- CN
- China
- Prior art keywords
- arable land
- sample set
- feature
- high resolution
- plant extraction
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Abstract
The invention belongs to remote sensing application technical field, tool is related to a kind of region adaptivity plant extraction method based on high resolution image, comprising: pre-processes to data;Multi-scale division is carried out to high resolution image, determines the optimum segmentation scale in arable land, obtains the optimum segmentation layer of arable land classification;Multiple arable land sample sets are constructed in optimum segmentation layer, form the attributive character space of each arable land sample set;Carry out optimal image feature spatial choice;The Fuzzy Classifier of plant extraction is constructed according to optimal image feature space;Plant extraction is carried out based on arable land sample set and Fuzzy Classifier.The present invention realizes the automatic identification of arable land object in such a way that sample set of ploughing is combined with Fuzzy Classifier.By the attributive character space of selection optimum segmentation scale and different arable land sample sets, the homogenieity of the cutting object of high resolution image and the extraction accuracy of Fuzzy Classifier are improved;By the building for sample set of ploughing, limitation requirement of the data source to phase sum number amount is reduced.
Description
Technical field
The invention belongs to remote sensing application technical fields, and in particular to a kind of region adaptivity based on high resolution image is cultivated
Ground extracting method.
Background technique
Quick, the accurate extraction of farmland information is to cultivated land dynamic change monitoring, Soil fertility investigation of cultivated land and evaluation, cultivated land protection and base
The delimitation of this farmland, land resources utilization degree analyzing, precision agriculture etc. are of great significance.
High resolution image enhances the internal diversity in arable land, so that diversity is presented in the spectrum performance of arable land covering,
Increase the difficulty that arable land is accurately extracted.At present there are many kinds of high resolution image plant extraction methods, have the following problems:
1) general plant extraction method is the growth characteristic according to crops, and when use is consistent more phase shadows of growth cycle
As analyzing and researching, the method is limited on promoting by the phase and quantity of data source;
2) using the plant extraction method of object-oriented using wide, but its that there are application regions is small, assorting process is multiple
Miscellaneous, the degree of automation is low, does not have universality and generalization.
Summary of the invention
The plant extraction method based on high resolution image that the object of the present invention is to provide a kind of, the plant extraction method
It is not limited by the phase of data source and quantity, and there is universality and generalization.
To achieve the above object, the invention proposes a kind of region adaptivity plant extraction side based on high resolution image
Method, comprising:
S1 pre-processes data;
S2 carries out multi-scale division to high resolution image, determines the optimum segmentation scale in arable land, obtains arable land classification
Optimum segmentation layer;
S3 constructs multiple arable land sample sets in optimum segmentation layer, forms the attributive character space of each arable land sample set;
S4 carries out optimal image feature spatial choice to each sample set of ploughing;
S5 constructs the Fuzzy Classifier of plant extraction according to optimal image feature space to each sample set of ploughing;
S6 carries out plant extraction based on arable land sample set and Fuzzy Classifier:
S61 carries out attribute to the object after step S2 optimum segmentation one by one according to the attributive character space of arable land sample set
Feature calculation;
S62, the attributive character for the arable land sample set that the obtained attributive character calculated result of S61 and step S3 are constructed respectively
Spatial dimension value is matched, if the attributive character value of object is remembered within the scope of the attributive character of a certain arable land sample set
It is 1, is otherwise denoted as 0, obtains matching matrix Dij
Wherein, l, p, n, b ..., w are respectively the attributive character of arable land sample set, and i is the number of arable land sample set, and j is to belong to
The number of property feature, the transposition of T representing matrix;
S63, according to weighted value shared by each attributive character in the arable land sample set of definition and matching matrix Dij, calculate separately
The similarity of object and each arable land sample set, obtains similarity matrix Si,
Wherein, δiFor sample set attributive character weight matrix, δiIt is carried out according to areas case customized;
S64 sets the threshold decision condition of similarity matrix Si, if the similarity of object and some arable land sample set meets
Threshold decision condition, then object matches with some arable land sample set;
S65 executes the corresponding Fuzzy Classifier of arable land sample set to match to the object according to the matching result of S64,
Class categories function E (i) and classification degree of membership ω (i) are obtained,
S66 carries out arable land weighted calculation using weighted voting algorithm, and calculation formula is as follows:
Wherein, E (x) is arable land weighting discriminant function, and n is the Fuzzy Classifier sum to match with object x, and i is fuzzy
The number of classifier, i=1 ..., n;
Whether S67 is that arable land judges that Rule of judgment is as follows to object:
Further, data prediction includes handling high resolution image and digital elevation model, wherein at image
Reason includes that radiation and geometric correction are carried out to image, and digital elevation model processing includes generating the gradient.
Further, in the step S2, the optimum segmentation scale in arable land is while meeting mean variance to be that peak value, shape refer to
Number variance is the segmentation scale of valley;
Wherein, mean variance SC 2Calculation formula be
Wherein, CLFor single object L-band luminance mean value,It is objects all in image in the bright of L-band
Spend mean value, m be image in object number summation, i be object number, i=1 ... m;
Shape index variance SSI 2Calculation formula be
Wherein, SI is the shape index value of single object,For the shape index mean value of objects all in image, m is shadow
The summation of object number as in, i are object number, i=1 ... m, b are the object perimeter of single object, and V is single object
Object volume.
Further, in the step S3,
The step of building arable land sample set, is as follows:
S31, by visual observation acquisition arable land sample, it is desirable that the sample of acquisition is uniformly distributed and representative, is covered all
The farmland types in district, quantity are no more than the 5% of whole region object number;
S32 carries out arable land sample according to arable land sample region, terrain slope feature, phenology feature and image feature
Classification, construct multiple arable land sample sets, and formed it is each arable land sample set attributive character space;
S33 rejects arable land sample:
If the average value of certain classification arable land sample each object feature isStandard deviation is
Std=(std1, std2, std3... ..., stdn), for a certain arable land sample X=(x1, x2, x3..., xn), if The rejecting of arable land sample is carried out with 2 times of standard deviations.
Further: the arable land sample region, terrain slope feature, phenology feature and image feature are respectively sample
This district title, the gradient, NDVI and brightness, the attributive character space of each arable land sample set include the district title of sample,
The gradient, NDVI and brightness.
Further, in the step S4, the step of optimal image feature spatial choice, is as follows:
S41, characteristics of objects space primary election of ploughing:
Arable land characteristics of objects space library is established for the spectrum, texture, shape feature of high resolution image:
S42 carries out optimal characteristics spatial choice based on feature variance:
The variance of each arable land each feature of sample set is calculated, the smallest 2~3 features of statistical variance are formed corresponding each
The optimal characteristics space of arable land sample set;
Feature variance SA 2Calculation formula is as follows:
Wherein, CAFor this feature value of single sample,For it is same arable land sample set in all samples this feature mean value,
M thus in sample set sample number, i=1 ... m;
Feature variance is compared and analyzed, the optimal characteristics space of each arable land sample set is selected.
Further, when constructing the Fuzzy Classifier of plant extraction in the step S5, according to obtained different arable land samples
The optimal characteristics space of collection, carries out feature combination by many experiments, carries out plant extraction experiment, structure using Fuzzy classification
Build multiple Fuzzy Classifiers of plant extraction.
Further: the Fuzzy Classifier of i-th of sample set f (i) is as follows:
Wherein, Max.diff is maximum heterogeneous, and NDVI is normalized differential vegetation index, and Brightness is brightness,
GLCM_ContrastFor texture contrast, Blue_Mean is blue wave band mean value, and Slope is the gradient.
The present invention realizes the automatic identification of arable land object in such a way that sample set of ploughing is combined with Fuzzy Classifier.
By the attributive character space of selection optimum segmentation scale and different arable land sample sets, the segmentation pair of high resolution image is improved
The homogenieity of elephant and the extraction accuracy of Fuzzy Classifier;By the building for sample set of ploughing, data source is reduced to phase sum number
The limitation requirement of amount;Using arable land sample set attributive character and Fuzzy Classifier be subordinate to angle value improve image from be divided into classification
The degree of automation, do not limited by region and data source, have generality, be conducive to promote.
Detailed description of the invention
Fig. 1 is that the present invention is based on the flow charts of the region adaptivity plant extraction method of high resolution image;
Fig. 2 is mean variance in embodiment with cutting scale change curve;
Fig. 3 is shape index variance yields in embodiment with cutting scale change curve.
Specific embodiment
Below by taking Hebei province as an example, in conjunction with attached drawing, the present invention is described in further detail.
Embodiment 1
A kind of region adaptivity plant extraction method based on high resolution image, as shown in Figure 1, comprising:
S1 pre-processes data
Select high resolution image for the multispectral and panchromatic remote sensing image of several SPOT6 of Hebei province, major phase is October
Part, wherein multispectral image (including blue, green, red and close red four wave bands) has 6m spatial resolution, and panchromatic remote sensing image is empty
Between resolution ratio be 1.5m, overshoot, geometric correction and visual fusion are carried out to the high resolution image, if it is single width high-resolution
Rate image does not need then to carry out visual fusion.
Meanwhile Selecting research area 30mDEM (digital elevation model) carry out subsidiary classification, to the digital elevation model into
The row gradient is generated, projective transformation and is cut processing (nonessential processing), guarantees its projection and piecemeal size and SPOT6 data one
It causes.
S2 carries out multi-scale division to high resolution image, determines the optimum segmentation scale in arable land
(1) multi-scale division is carried out to high resolution image
Multi-scale division is executed segmentation under the partitioning parameters of setting and is appointed using heterogeneous the smallest region merging algorithm
Business, by given scale threshold value, according to the homogeneity criterion of color and shape, based on the heterogeneous the smallest principle of object, by light
The similar adjacent picture elements of spectrum information merge into significant object.
The merging in region, formula are split using multiple features weighted criterion are as follows:
F=w1×Hcolor+w2×Hshape+w3×Htexture
Wherein, F is heterogeneity, w1、w2And w3The respectively feature weight of spectrum, shape and texture, w1+w2+w3=1;
Hcolor、HshapeAnd HtextureThe respectively characteristic value of spectrum, shape and textural characteristics.
Easy health software is utilized to carry out image in conjunction with expertise and visual observation by pretreated high resolution image
Multi-scale division, wherein selected spectrum and shape criteria ratio is 9: 1, (the two is all for smoothness parameter and compact degree parameter
Textural characteristics) ratio be 1: 1, select 15 segmentation scale threshold values (10,20,30,40,50,60,70,80,90,100,
110,120,130,140,150) multi-scale division is carried out, the imaged object layer of different scale is formd.Multi-scale division parameter
Selection principle be: under the premise of meeting necessary shape criteria, color standard is used as far as possible, this is because in image most
It is important that spectral information, the weight of shape criteria is too high, can reduce the quality of segmentation instead.If the boundary ratio of ground object target
It is smoother, just set higher smoothness parameter, the type of ground objects compact for shape, it should which biggish compact degree ginseng is set
Number.
(2) the optimum segmentation scale for determining arable land, obtains the optimum segmentation layer of arable land classification;
Suitable scale selection is defined so that the inside heterogeneity of imaged object after dividing is as small as possible, and different
Heterogeneity between class is big as far as possible.
The present embodiment uses a kind of optimal scale calculation method combined based on object mean variance and shape index variance,
Its principle is to comprehensively consider the spectral signature and shape feature characteristic in arable land, counts all object wave band mean values in different dividing layers
With the variance of shape index, then respectively using the mean variance of object and shape index variance as Y-axis, Image Segmentation scale is X
Axis, rendered object mean variance and shape index variance are with the curve for dividing dimensional variation, from two curve distribution forms
Determine the optimum segmentation scale in arable land.
Mean variance SC 2Calculation formula be
It wherein, is mean variance, CLFor single object L-band luminance mean value,Exist for objects all in image
The luminance mean value of L-band, m be image in object number summation, i be object number, i=1 ... m.
Shape index variance SSI 2Calculation formula be
It wherein, is shape index variance, SI is the shape index value of single object,For the shape of objects all in image
Mean value of index, m be image in object number summation, i be object number, i=1 ... m, b be single object object perimeter,
V is the object volume of single object.
15 multi-scale division layers calculate the mean variance and shape index of SPOT6 high resolution image near infrared band
Variance it is as shown in Figure 2.
The features such as arable land has spectral signature relatively uniform, and regular shape/pattern of farming is more regular, according to these
Feature, the reference value of the optimum segmentation in arable land occur to intersect in the partition value of mean variance peak value and shape index variance valley
Place, i.e., meeting the segmentation scale that mean variance is peak value, shape index variance is valley simultaneously is arable land optimum segmentation scale.This
It is that the spectral differences opposite sex because when pure object increases, between adjacent object increases, and when blending objects increase, with phase adjacency pair
Spectrum heterogeneity as between reduces, and mean variance becomes smaller;And shape index is the smoothness for characterizing subject surface, shape is more advised
Then, the boundary light slippage opposite sex between adjacent object is smaller, and the variance of shape index just becomes smaller.
Obtained by Fig. 2 tracing analysis, arable land optimum segmentation scale be 70, by visual observation it can be seen that on 70 scales compared with
Good carries out arable land optimal segmentation.
All image processing below are carried out on the optimum segmentation layer that arable land optimum segmentation scale is 70.
S3 constructs multiple arable land sample sets in optimum segmentation layer, forms the attributive character space of each arable land sample set
In view of same class atural object due to its size, distribution, density are different and cause difference in spectral response often
It is difficult to will lead to classification, thus by a certain atural object according to its spectrum and other features be subdivided into several subclasses carry out classification again can be with
It is effectively reduced classification difficulty.
The building of arable land sample set is carried out using following steps:
(1) arable land sample is acquired by visual observation first, it is desirable that the sample of acquisition is uniformly distributed and representative, basic culvert
The farmland types in all districts are covered, quantity is no more than the 5% of whole region object number.
(2) according to arable land sample region, terrain slope feature, phenology feature and image feature, arable land sample is carried out
Classification, construct multiple arable land sample sets, and formed it is each arable land sample set attributive character space.
(3) arable land sample is rejected:
Although during ploughing samples selection, it then follows uniformly, rules, the analysis such as homogeneity find, a small amount of sample
The characteristics of objects variation on ground is too big, therefore standard deviation method is used to reject arable land sample.
If the average value of certain classification arable land sample each object feature isStandard deviation is
Std=(std1, std2, std3... ..., stdn), for a certain arable land sample X=(x1, x2, x3..., xn), if The rejecting of arable land sample is carried out with 2 times of standard deviations.
The district title of sample, the gradient (Slope), normalized differential vegetation index (NDVI) and bright are had chosen by above method
Degree (Brightness) four indexs carry out the building of arable land sample set to test block, generate four classes arable land sample set, sample
The attributive character space of this collection is as shown in the table:
S4 carries out optimal image (arable land object) feature space selection to each sample set of ploughing
(1) arable land characteristics of objects space primary election
The features such as spectrum, texture, shape for high resolution image establish arable land characteristics of objects space library, such as following table institute
Show:
Feature classification | Characteristic parameter |
Spectral signature | Mean value, brightness, normalized differential vegetation index, soil vegetative cover index |
Shape feature | Length-width ratio, density, compact degree, the gradient |
Textural characteristics | Maximum heterogeneity, homogeneity degree, contrast, entropy |
(2) based on the optimal characteristics spatial choice of feature variance
In the characteristics of objects space primary election result of arable land, selection effective cultivated land trait composition plant extraction less as far as possible
Optimal characteristics space.It is each by calculating each arable land sample set using the optimal characteristics spatial choice method based on feature variance
The variance of a feature, the smallest 2~3 features of statistical variance form the optimal characteristics space of corresponding each arable land sample set.Feature
Variance calculation formula is as follows:
Wherein, SA 2For the variance of a certain feature, CAFor this feature value of single sample,For institute in same arable land sample set
Have this feature mean value of sample, m thus in sample set sample number, i=1 ... m.
By the comparative analysis to feature variance, the optimal characteristics space of each arable land sample set is selected, is as follows:
S5 constructs the Fuzzy Classifier of plant extraction according to optimal image feature space to each sample set of ploughing
According to the optimal characteristics space of obtained different arable land sample sets, feature combination is carried out by many experiments, is used
Fuzzy classification carries out plant extraction experiment, finally constructs multiple Fuzzy Classifiers of plant extraction.
The fuzzy membership functions of Fuzzy Classifier uses the determination method based on Bayes criterion, is based on feature space root
Determine that imaged object is under the jurisdiction of certain a kind of degree according to ambiguity function, it is not only to differentiate classification institute according to feature space
Belong to, while according to the subjection degree threshold value for dividing ground class, judging the accuracy of ground class again.
In formula: P ' (g) is the prior probability that g class occurs,
For g class fuzzy vector mean value,Fuzzy covariance matrix, m are the sum of object, and X is the object of classification, and i is pair
The number of elephant, i=1,2 ... ..., G, G are the total number of g class object.
By many experiments, four corresponding Fuzzy Classifiers of arable land sample set are obtained, are respectively as follows:
Wherein, f (i) is the Fuzzy Classifier of i-th of sample set.
S6 carries out plant extraction based on arable land sample set and Fuzzy Classifier
S61, according to the attributive character space of arable land sample set, after the optimum segmentation layer segmentation obtained one by one to step S2
Object x (x=1,2 ... n wherein n be object total number) carries out attributive character calculating, including district title, the gradient, NDVI and bright
Degree;
S62, the attributive character for the arable land sample set that the obtained attributive character calculated result of S61 and step S3 are constructed respectively
Spatial dimension value is matched, if the attributive character value of object is remembered within the scope of the attributive character of a certain arable land sample set
It is 1, is otherwise denoted as 0, obtains matching matrix Dii
Wherein, l, p, n, b ..., w are respectively the attributive character of arable land sample set, such as l, p, and n, b respectively represent district name
Claim, the gradient, NDVI and brightness, i is the number of arable land sample set, and j is the number of attributive character, the transposition of T representing matrix;
S63, according to weighted value shared by each attributive character in the arable land sample set of definition and matching matrix Dij, calculate separately
The similarity of object and each arable land sample set, obtains similarity matrix Si,
Wherein, δiFor sample set attributive character weight matrix, δiIt is customized according to areas case progress, under normal circumstances, δi
It is 0.5.
In the present embodiment, j=4, the district title in sample set attributive character, the gradient, the weight difference of NDVI and brightness
It is 0.3,0.2,0.3,0.2,
S64 sets similarity matrix SiThreshold decision condition, if the similarity of object x and some arable land sample set is full
Sufficient threshold decision condition, then object matches with some arable land sample set;
Threshold condition is set in the present embodiment are as follows:
If Si>=0.5 item object x matches with j-th of arable land sample set.
Otherwise, restart step S61, calculate next object x+1.
Wherein, object x can match with multiple sample sets simultaneously.
S65 executes the corresponding Fuzzy Classifier of arable land sample set to match to the object according to the matching result of S64,
Class categories function E (i) and classification degree of membership are obtained, principle is to set 1 for multiple Fuzzy Classifiers execution classification results
With 0, wherein 1 is arable land, 0 is bare place, and Fuzzy Classifier arable land is then subordinate to angle value as weighted value, using weighting throwing
Ticket rule is judged, is met the final of certain threshold range condition and is determined as ploughing.
S66 carries out arable land weighted calculation using weighted voting algorithm, and calculation formula is as follows:
Wherein, E (x) is arable land weighting discriminant function, and n is the Fuzzy Classifier sum to match with object x, and i is fuzzy
The number of classifier, i=1 ..., n;
Whether S67 is that arable land judges that Rule of judgment is as follows to object:
Repetitive cycling, until whole calculation and objects are completed.
Claims (8)
1. a kind of region adaptivity plant extraction method based on high resolution image, which is characterized in that the extracting method packet
It includes:
S1 pre-processes data;
S2 carries out multi-scale division to high resolution image, determines the optimum segmentation scale in arable land, obtains the optimal of arable land classification
Dividing layer;
S3 constructs multiple arable land sample sets in optimum segmentation layer, forms the attributive character space of each arable land sample set;
S4 carries out optimal image feature spatial choice to each sample set of ploughing;
S5 constructs the Fuzzy Classifier of plant extraction according to optimal image feature space to each sample set of ploughing;
S6 carries out plant extraction based on arable land sample set and Fuzzy Classifier:
S61 carries out attributive character to the object after step S2 optimum segmentation one by one according to the attributive character space of arable land sample set
It calculates;
S62, the attributive character space for the arable land sample set that the obtained attributive character calculated result of S61 and step S3 are constructed respectively
Value range is matched, if the attributive character value of object is denoted as 1 within the scope of the attributive character of a certain arable land sample set,
Otherwise it is denoted as 0, obtains matching matrix Dij
Wherein, l, p, n, b ..., w are respectively the attributive character of arable land sample set, and i is the number of arable land sample set, and j is that attribute is special
The number of sign, the transposition of T representing matrix;
S63, according to weighted value shared by each attributive character in the arable land sample set of definition and matching matrix Dij, calculate separately object
With the similarity of each arable land sample set, similarity matrix S ' is obtained,
Wherein, δ is sample set attributive character weight matrix, and δ carries out customized according to areas case;
S64 sets the threshold decision condition of similarity matrix S, if the similarity of object and some arable land sample set meets threshold value
Rule of judgment, then object matches with some arable land sample set;
S65 executes the corresponding Fuzzy Classifier of arable land sample set to match to the object, obtains according to the matching result of S64
Class categories function E (k) and classification degree of membership ω (k),
S66 carries out arable land weighted calculation using weighted voting algorithm, and calculation formula is as follows:
Wherein, E (x) is arable land weighting discriminant function, and n is the Fuzzy Classifier sum to match with object x, and k is fuzzy classification
The number of device, k=1 ..., n;
Whether S67 is that arable land judges that Rule of judgment is as follows to object:
2. the region adaptivity plant extraction method based on high resolution image as described in claim 1, which is characterized in that number
Data preprocess includes handling high resolution image and digital elevation model, wherein image processing includes carrying out to image
Radiation and geometric correction, digital elevation model processing include generating the gradient.
3. the region adaptivity plant extraction method based on high resolution image as described in claim 1, which is characterized in that institute
It states in step S2, the optimum segmentation scale in arable land is while meeting point that mean variance is peak value, shape index variance is valley
Cut scale;
Wherein, mean variance SC 2Calculation formula be
Wherein, CLhFor single object L-band luminance mean value,It is equal in the brightness of L-band for objects all in image
Value, m be image in object number summation, h be object number, h=1 ... m;
Shape index variance SSI 2Calculation formula be
Wherein, SInFor the shape index value of single object,For the shape index mean value of objects all in image, m is in image
The summation of object number, h be object number, h=1 ... m, b be single object object perimeter, V be single object object
Volume.
4. the region adaptivity plant extraction method based on high resolution image as described in claim 1, it is characterised in that: institute
The step of stating in step S3, constructing arable land sample set is as follows:
S31, by visual observation acquisition arable land sample, it is desirable that the sample of acquisition is uniformly distributed and representative, covers all districts
Farmland types, quantity be no more than whole region object number 5%;
S32 carries out returning for arable land sample according to arable land sample region, terrain slope feature, phenology feature and image feature
Class constructs multiple arable land sample sets, and forms the attributive character space of each arable land sample set;
S33 rejects arable land sample:
If the average value of certain classification arable land sample each object feature isStandard deviation is Std=
(std1, std2, std3... ..., stdn), for a certain arable land sample X=(x1, x2, x3..., xn), if The rejecting of arable land sample is carried out with 2 times of standard deviations.
5. the region adaptivity plant extraction method based on high resolution image as claimed in claim 4, it is characterised in that: institute
State arable land sample region, terrain slope feature, phenology feature and image feature be respectively the district title of sample, the gradient,
The attributive character space of NDVI and brightness, each arable land sample set includes district title, the gradient, NDVI and the brightness of sample.
6. the region adaptivity plant extraction method based on high resolution image as described in claim 1, it is characterised in that: institute
It states in step S4, the step of optimal image feature spatial choice is as follows:
S41, characteristics of objects space primary election of ploughing:
Arable land characteristics of objects space library is established for the spectrum, texture, shape feature of high resolution image:
S42 carries out optimal characteristics spatial choice based on feature variance:
The variance of each arable land each feature of sample set is calculated, the smallest 2~3 features of statistical variance form corresponding each arable land
The optimal characteristics space of sample set;
Feature variance SA 2Calculation formula is as follows:
Wherein, CA is the characteristic value of single sample,For the characteristic mean of all samples in same arable land sample set, p is ploughed thus
The number of sample in ground sample set, p=1 ... q;
Feature variance is compared and analyzed, the optimal characteristics space of each arable land sample set is selected.
7. the region adaptivity plant extraction method based on high resolution image as described in claim 1, it is characterised in that: institute
When stating the Fuzzy Classifier for constructing plant extraction in step S5, according to the optimal characteristics space of obtained different arable land sample sets,
Feature combination is carried out by many experiments, plant extraction experiment is carried out using Fuzzy classification, constructs the multiple of plant extraction
Fuzzy Classifier.
8. the region adaptivity plant extraction method based on high resolution image as claimed in claim 7, it is characterised in that: the
The Fuzzy Classifier of i sample set f (i) is as follows:
Wherein, Max.diff is maximum heterogeneous, and NDVI is normalized differential vegetation index, and Brightness is brightness,
GLCM_ContrastFor texture contrast, Blue_Mean is blue wave band mean value, and Slope is the gradient.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201511034394.3A CN105608474B (en) | 2015-12-30 | 2015-12-30 | Region adaptivity plant extraction method based on high resolution image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201511034394.3A CN105608474B (en) | 2015-12-30 | 2015-12-30 | Region adaptivity plant extraction method based on high resolution image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105608474A CN105608474A (en) | 2016-05-25 |
CN105608474B true CN105608474B (en) | 2019-07-09 |
Family
ID=55988398
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201511034394.3A Active CN105608474B (en) | 2015-12-30 | 2015-12-30 | Region adaptivity plant extraction method based on high resolution image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105608474B (en) |
Families Citing this family (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106548141B (en) * | 2016-11-01 | 2019-07-26 | 南京大学 | A kind of object-oriented farmland information extraction method based on the triangulation network |
CN106446875A (en) * | 2016-11-16 | 2017-02-22 | 航天恒星科技有限公司 | County remote-sensing scale-oriented crop planting area information extraction method and device |
CN106767834B (en) * | 2017-01-24 | 2019-11-08 | 哈尔滨工程大学 | A kind of AUV underwater terrain matching critical subset segmentation method based on fuzzy entropy |
CN107194942B (en) * | 2017-03-27 | 2020-11-10 | 广州地理研究所 | Method for determining image classification segmentation scale threshold |
CN107577998B (en) * | 2017-08-21 | 2019-02-26 | 北京阿克西斯信息技术有限公司 | A kind of automatic identification Agricultural land system, implementation method |
CN108090567B (en) * | 2018-01-19 | 2021-07-27 | 国家电网公司 | Fault diagnosis method and device for power communication system |
CN108629494B (en) * | 2018-04-19 | 2020-11-06 | 三峡大学 | Drought level assessment method and system |
CN109522788B (en) * | 2018-09-30 | 2020-11-06 | 广州地理研究所 | City range extraction method and device based on random forest classification algorithm and electronic equipment |
CN109635827A (en) * | 2018-12-21 | 2019-04-16 | 长光卫星技术有限公司 | Based on the sugarcane plot extracting method for containing red side spectral coverage high-resolution satellite image |
CN109816707A (en) * | 2018-12-25 | 2019-05-28 | 中铁第四勘察设计院集团有限公司 | A kind of field of opencast mining information extracting method based on high-resolution satellite image |
CN109740570B (en) * | 2019-01-18 | 2020-01-03 | 中国水利水电科学研究院 | Terrace remote sensing identification method based on terrain index and climate difference |
CN111476182B (en) * | 2020-04-13 | 2021-06-01 | 中国科学院空天信息创新研究院 | Terrace extraction method based on multi-source data and multi-directional texture filtering analysis |
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 |
CN111950530B (en) * | 2020-09-08 | 2024-04-12 | 中国水利水电科学研究院 | Multi-feature optimization and fusion method for crop planting structure extraction |
CN113505948B (en) * | 2021-09-13 | 2021-11-19 | 四川师范大学 | Future ecological security pattern prediction and optimization method based on Bayesian network |
CN116188880B (en) * | 2023-05-05 | 2023-07-18 | 中国科学院地理科学与资源研究所 | Cultivated land classification method and system based on remote sensing image and fuzzy recognition |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5361309A (en) * | 1989-09-07 | 1994-11-01 | Canon Kabushiki Kaisha | Character recognition apparatus and method with low-resolution storage for character extraction |
CN101706950A (en) * | 2009-07-13 | 2010-05-12 | 中国科学院遥感应用研究所 | High-performance implementation method for multi-scale segmentation of remote sensing images |
CN102496151A (en) * | 2011-12-08 | 2012-06-13 | 南京大学 | Method for multi-scale segmentation of high-resolution remote sensing images |
CN104463168A (en) * | 2014-11-25 | 2015-03-25 | 中国科学院地理科学与资源研究所 | Automatic waste empty house site information extraction method based on remote-sensing image |
CN104881677A (en) * | 2015-05-08 | 2015-09-02 | 北京师范大学 | Optimum segmentation dimension determining method for remotely-sensed image land cover classification |
-
2015
- 2015-12-30 CN CN201511034394.3A patent/CN105608474B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5361309A (en) * | 1989-09-07 | 1994-11-01 | Canon Kabushiki Kaisha | Character recognition apparatus and method with low-resolution storage for character extraction |
CN101706950A (en) * | 2009-07-13 | 2010-05-12 | 中国科学院遥感应用研究所 | High-performance implementation method for multi-scale segmentation of remote sensing images |
CN102496151A (en) * | 2011-12-08 | 2012-06-13 | 南京大学 | Method for multi-scale segmentation of high-resolution remote sensing images |
CN104463168A (en) * | 2014-11-25 | 2015-03-25 | 中国科学院地理科学与资源研究所 | Automatic waste empty house site information extraction method based on remote-sensing image |
CN104881677A (en) * | 2015-05-08 | 2015-09-02 | 北京师范大学 | Optimum segmentation dimension determining method for remotely-sensed image land cover classification |
Non-Patent Citations (4)
Title |
---|
多尺度分割技术在高分辨率影像信息提取中的应用研究;孙波中;《中国优秀硕士学位论文全文数据库 信息科技辑》;20120115(第01期);全文 |
多特征融入的自适应遥感影像多尺度分割;周亚男等;《武汉大学学报信息科学版》;20130131;第38卷(第1期);全文 |
面向对象的高分辨率遥感影像信息提取_以耕地提取为例;李敏等;《遥感应用》;20080630;全文 |
高分辨率遥感影像耕地分层提取方法;陈杰等;《农业工程学报》;20150227;第31卷(第3期);全文 |
Also Published As
Publication number | Publication date |
---|---|
CN105608474A (en) | 2016-05-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105608474B (en) | Region adaptivity plant extraction method based on high resolution image | |
CN111598045B (en) | Remote sensing farmland change detection method based on object spectrum and mixed spectrum | |
CN104598908A (en) | Method for recognizing diseases of crop leaves | |
CN112381013B (en) | Urban vegetation inversion method and system based on high-resolution remote sensing image | |
Cai et al. | Residual-capsule networks with threshold convolution for segmentation of wheat plantation rows in UAV images | |
CN107832797B (en) | Multispectral image classification method based on depth fusion residual error network | |
Usman | Satellite imagery land cover classification using k-means clustering algorithm computer vision for environmental information extraction | |
CN109446986B (en) | Effective feature extraction and tree species identification method for tree laser point cloud | |
CN108280396A (en) | Hyperspectral image classification method based on depth multiple features active migration network | |
CN107330875A (en) | Based on the forward and reverse heterogeneous water body surrounding enviroment change detecting method of remote sensing images | |
CN109034233A (en) | A kind of high-resolution remote sensing image multi classifier combination classification method of combination OpenStreetMap | |
Chen et al. | Agricultural remote sensing image cultivated land extraction technology based on deep learning | |
CN114694036A (en) | High-altitude area crop classification and identification method based on high-resolution images and machine learning | |
Zaabar et al. | Application of convolutional neural networks with object-based image analysis for land cover and land use mapping in coastal areas: A case study in Ain Témouchent, Algeria | |
Jónsson | RGB and Multispectral UAV image classification of agricultural fields using a machine learning algorithm | |
CN109740504A (en) | A method of sea area resources are extracted based on remote sensing image | |
Babu | Optimized performance and utilization analysis of real-time multi spectral data/image categorization algorithms for computer vision applications | |
CN102622345B (en) | High-precision land-utilization remote sensing updating technology with synergistic multisource spatio-temporal data | |
CN115953685A (en) | Multi-layer multi-scale division agricultural greenhouse type information extraction method and system | |
Mitrakis et al. | A novel self‐organizing neuro‐fuzzy multilayered classifier for land cover classification of a VHR image | |
CN115063678A (en) | Soybean planting area extraction method based on domestic GF-6WFV data | |
CN114022782A (en) | Sea fog detection method based on MODIS satellite data | |
Akoso | Remotely Sensed Image Classification: Supervised Classification Algorithm Using Envi 5.0 Software | |
CN113240051B (en) | Cultivated land block classification method based on genetic programming algorithm | |
CN116310810B (en) | Cross-domain hyperspectral image classification method based on spatial attention-guided variable convolution |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |