Invention content
The present invention the technical problem to be solved is that provide a kind of accurate extraction cotton planting information of automation based on
The cotton remote-sensing monitoring method of phenology analysis.
To achieve the above object, present invention proposition is included the following steps based on the cotton remote-sensing monitoring method that phenology is analyzed:
A kind of cotton remote-sensing monitoring method based on phenology analysis, includes the following steps:
Step 1: the analysis of calmodulin binding domain CaM chief crop phenology and cotton Spectral Characteristics Analysis, specify cotton remote sensing monitoring
Crucial phase;
1.1, cotton planting phase main interference crop type and its phenology are determined, makes phenology table, described is main
It includes winter wheat, corn, peanut, soybean to interfere crops;
1.2, cotton spectral signature curve is drawn;
1.3, in conjunction with the cotton spectral signature tracing analysis of different stages of growth, and it is based on remote sensing image visual interpretation, built
The remote sensing image interpretation mark in vertical cotton growth stage;
Detailed process is as follows:
Cotton is in sowing time and seeding stage, and bare area information is presented in remote sensing image;Cotton is in flower bud phase, and remote sensing image is in
The terrain surface specifications at cotton vegetation initial stage are presented in pink;Cotton is in flowering and boll-setting period and the term of opening bolls, and remote sensing image is in rose, is in
The luxuriant vegetation terrain surface specifications of existing cotton;Cotton is in harvest time, and remote sensing image is in light red, the withered vegetation of cotton is presented
Region feature;
1.4, gradient divides;
The cotton identification critical period is divided into four gradients, first gradient be do not sow, sowing time and seedling stage, topographical features are
Bare area feature;Second gradient is flower bud phase, flowering and boll-setting period, the term of opening bolls, and topographical features are vegetation characteristics;3rd gradient is ripe harvests
Phase, topographical features are vegetation characteristics;After 4th gradient is ripe harvest time, topographical features are bare area feature;
1.5, the crucial phase of cotton identification in the gradient is chosen using Moving split-window technique in each gradient scope;
Detailed process is as follows:
In each gradient, using the ten days moving window in phenology table, interference agriculture opposite with cotton topographical features as step-length
Phase when crop species number maximum is determined as the crucial phase of cotton identification in the gradient;
1.6, in conjunction with the availability of remotely-sensed data, cotton remote sensing prison is determined in the crucial phase of cotton identification in each gradient
Survey crucial phase.
Step 2: flowering and boll-setting period residing for cotton and choosing a phase vegetation growth vigorous period between the phase in the term of opening bolls
Remote sensing image;
Step 3: obtaining crucial phase remote sensing image, pre-processed;
It is reference with the remote sensing image of a phase, image registration is carried out to the remote sensing image of other phases;
Step 4: carrying out multi-scale division to multi-temporal remote sensing image, layering structure grader tentatively extracts cotton;
Detailed process is as follows:
4.1, the remote sensing image of crucial phase, which carries out multi-scale division, to be identified to multiple cottons, optimum segmentation scale is simultaneously
The spectrum Split Index for meeting the remote sensing image object of each phase each phase corresponding when being peak value divides the flat of scale-value
Mean value;
4.2, different image information layer L is builtt, t is different images issue used;
4.2.1 the remote sensing in vegetation growth vigorous period, is chosen between the flowering and boll-setting period residing for cotton and the phase in the term of opening bolls
Image;It is denoted as L to the 1st layer of the remote sensing image structure classification layer in the vegetation growth of selection vigorous period1, in classification layer L1In, it builds
Vertical multidate spectral signature grader f (L11), vegetation atural object all in remote sensing image is extracted, the vegetation atural object includes garden
Forest land, cotton, interference crops;In classification layer L1In, build multidate spectral signature grader f (L12), reject remote sensing image
In all landscape ground, remaining vegetation terrestrial object information includes cotton, interference crops in remote sensing image;
4.2.2 the remote sensing image of the crucial phase of first gradient cotton identification, is chosen, the 2nd layer of structure classification layer is denoted as L2,
In classification layer L2In, establish multidate spectral signature grader f (L2), reject the interference crops of spring riotous growth;
4.2.3 the remote sensing image of the crucial phase of the second gradient cotton identification, is chosen, the 3rd layer of structure classification layer is denoted as L3,
In classification layer L3In, establish multidate spectral signature grader f (L3), reject the interference crops corn, big that bare area feature is presented
Beans, remaining vegetation terrestrial object information is cotton and peanut in remote sensing image;
4.2.4 the remote sensing image of the crucial phase of 3rd gradient cotton identification, the 4th layer of L of structure classification layer, are chosen4, dividing
Class layer L4In, establish multidate spectral signature grader f (L4), reject the interference crops peanut that bare area feature is presented, remote sensing shadow
Remaining vegetation terrestrial object information is the preliminary information result of cotton as in.
In formula, VndviExponential quantity, V are normalized for vegetationBrightnessFor wave band average brightness value, VGLCMContrast(all dir)For
Texture comparison's angle value, VBlue_MeanFor blue wave band wave band average value, VMax.diffFor maximum heterogeneity index value, a, b, c, d, e, f,
G, h is threshold value.
Step 5: preliminary information result to cotton in the remote sensing image that is obtained in step 4 and may mix and be divided into cotton
Other interference secondary multi-scale divisions of crop information, build grader, finely extract Cotton information;
5.1, the preliminary information result to cotton in the remote sensing image that is obtained in step 4 and its for being divided into cotton may be mixed
He interferes crop information to carry out secondary multi-scale division;It is peanut, soybean with mixed point of interference crops of cotton;Described two
The remote sensing image that secondary multi-scale division is based on is the remote sensing image corresponding to the crucial phase of the last one gradient cotton identification;
Secondary multi-scale division choice of optimal scale:Secondary multi-scale division object is as target area, to interfere farming
The minimum remote sensing image near infrared band mean variance of object is assessment parameter, is segmentation scale lower limit with 3, and 2 be step-length, in step 4
The optimum segmentation scale of initial partitioning is the upper limit, and it is secondary multiple dimensioned point to select corresponding segmentation scale when mean variance value maximum
The optimum segmentation scale cut;
5.2, grader is built;
In conjunction with cotton spectral signature curve and textural characteristics structure grader f (L), it is shown below, from secondary more rulers
The interference crops of rejecting and mixed point of cotton in cotton preliminary information result after degree segmentation, 1 information of extraction cotton subclass, from it
He interferes 3 information of 2 information of cotton subclass and cotton subclass that extraction leakage divides respectively in crops peanut, soybean;
Wherein, VndviExponential quantity, V are normalized for vegetationBrightnessFor wave band average brightness value, VGLCM Contrast(all dir)
For texture comparison's angle value, VMean4For the wave band average value of near infrared band, k, l, m, n, o, p are threshold value;
Step 6: carrying out merger, output drawing to the cotton various information of extraction;
Further, in the step 1, cotton spectral signature curve construction step is as follows:
1) according to the sampling of Neuman (2011) rule, random selection cotton sample prescription Sj, j=1,2,3 ... n, n are
The cotton quadrat number of selection;
2) cotton growth stage p is calculated, cotton sample prescription S in the image wave band b of remote sensing imagejDN values average value or
The average value of spectrum,P ∈ [non-sowing time, sowing time, seedling stage, flower bud phase, flowering and boll-setting period, the term of opening bolls, ripe harvest time];
3) to all cotton sample prescription S in each cotton growth stage p, each image wave band bjDN values average value or light
The average value of spectrumMean value is taken again, obtains all cotton sample prescriptions in each cotton growth stage p, each image wave band b
Value, be denoted as
4) drafting abscissa is b, ordinate isP cotton spectral signature curve;Drafting abscissa is p, indulges and sit
It is designated asB cotton spectral signature curve.
The advantageous effect that is reached of the present invention is:
The present invention combines the hierarchical classification method based on remote sensing image by the phenological period analysis of cotton and interference crops
The remote sensing for realizing cotton planting information automatically extracts, by fully consider cotton the spectral signature of different growing stages and with
The SPECTRAL DIVERSITY and texture difference of other interference crops, make cotton planting information extraction result more accurately and quickly, are easy to
Promotion and application.
Specific implementation mode
When a kind of cotton remote-sensing monitoring method based on phenology analysis of the present invention shoots more suitable for same satellite
Phase remote sensing image, specific implementation flow as shown in Figure 1, include the following steps:
Step 1: the analysis of calmodulin binding domain CaM chief crop phenology and cotton Spectral Characteristics Analysis, specify cotton remote sensing monitoring
Crucial phase;
1.1, cotton planting phase main interference crop type and its phenology are determined, phenology table is made.
1.2, cotton indicatrix is drawn, determines cotton remote sensing monitoring key phase.
According to each growth phase p of cotton in phenology table, p ∈ [non-sowing time, sowing time, seedling stage, flower bud phase, flowering and boll-setting period, blow-of-cottons
Phase, ripe harvest time], remote sensing image is chosen, cotton spectral signature curve is drawn.Cotton growing stage spectral signature is according to Neuman
(2011) sampling rule draws two kinds of cotton spectrum that abscissa is growth period and different-waveband respectively using sample prescription mean value
Indicatrix.
Cotton spectral signature curve construction step is as follows:
1) according to the sampling of Neuman (2011) rule, random selection cotton sample prescription Sj, j=1,2,3 ... n, n are
The cotton quadrat number of selection.
2) cotton growth stage p is calculated, cotton sample prescription S in the image wave band b of remote sensing imagejDN values average value or
The average value of spectrum,P ∈ [non-sowing time, sowing time, seedling stage, flower bud phase, flowering and boll-setting period, the term of opening bolls, ripe harvest time].
3) to all cotton sample prescription S in each cotton growth stage p, each image wave band bjDN values average value or light
The average value of spectrumMean value is taken again, obtains all cotton sample prescriptions in each cotton growth stage p, each image wave band b
Value, be denoted as
4) it is b to draw abscissa, and ordinate isP cotton spectral signature curve;Drafting abscissa is p, indulges and sit
It is designated asB cotton spectral signature curve.
1.3, in conjunction with the cotton spectral signature tracing analysis of different stages of growth, and it is based on remote sensing image visual interpretation, built
The remote sensing image interpretation mark in vertical cotton growth stage;Detailed process is as follows:
Cotton is in sowing time and seeding stage, and bare area information is presented in remote sensing image;Cotton is in flower bud phase, in growth
Primary stage, vegetation can cover earth's surface, and the terrain surface specifications at cotton vegetation initial stage are presented in remote sensing image pinkiness;Cotton is in
Flowering and boll-setting period and the term of opening bolls, growth is vigorous, and remote sensing image is in rose, and the luxuriant vegetation terrain surface specifications of cotton are presented;Cotton is in
Harvest time, remote sensing image are in light red, and the withered vegetation terrain surface specifications of cotton are presented, match with spectrum analysis.
1.4, gradient divides;
The topographical features that spectral signature and interpretation mark based on cotton different stages of growth are presented, cotton is identified and is closed
The key phase is divided into four gradients, and four gradients divide and topographical features are as follows:
Gradient divides |
Phenological stage |
Topographical features |
First gradient |
It does not sow, sowing time and seedling stage |
Bare area feature |
Second gradient |
Flower bud phase, flowering and boll-setting period, the term of opening bolls |
Vegetation characteristics |
3rd gradient |
Ripe harvest time |
Vegetation characteristics |
4th gradient |
After ripe harvest time |
Bare area feature |
The determination method of crop species is as follows in each gradient:
Crop species in first gradient include cotton and all interference crops;
Second, third, the crop species in 4th gradient do not include in a upper gradient under the crucial phase of cotton identification with cotton
The opposite interference crops of flower topographical features.
1.5, the crucial phase of cotton identification in the gradient is chosen using Moving split-window technique in each gradient scope.
Detailed process is as follows:
In each gradient, using the ten days moving window in phenology table, interference agriculture opposite with cotton topographical features as step-length
Phase when crop species number maximum is determined as the crucial phase of cotton identification in the gradient.
1.6, it in conjunction with the availability of remotely-sensed data, is determined in each gradient in the crucial phase of cotton identification in each gradient
Cotton remote sensing monitoring key phase.
Step 2: the influence that crops are extracted in comprehensive forest land, increases the image in a phase vegetation growth vigorous period, improve
Cotton extraction accuracy, increased key flowering and boll-setting period of the phase image preferably residing for cotton and between term of opening bolls phase, in cotton
The remote sensing image in vegetation growth vigorous period is chosen between the phase of residing flowering and boll-setting period and the term of opening bolls.
Step 3: obtaining crucial phase remote sensing image, pre-processed;
It is reference with the remote sensing image of a phase, image registration is carried out to the remote sensing image of other phases.
Step 4: carrying out multi-scale division to multidate image, layering structure grader tentatively extracts cotton;
Detailed process is as follows:
4.1, the remote sensing image of crucial phase, which carries out multi-scale division, to be identified to multiple cottons, optimum segmentation scale is simultaneously
The spectrum Split Index for meeting the remote sensing image object of each phase each phase corresponding when being peak value divides the flat of scale-value
Mean value.
4.2, different image information layer L is builtt, t is different images issue used, and Fig. 2 gives cotton remote sensing monitoring
Hierarchical classification system;
4.2.1 the 1st layer of the remote sensing image, based on vegetation growth vigorous period, structure classification layer is denoted as L1, in classification layer
L1In, establish multidate spectral signature grader f (L11), extract all vegetation atural object in remote sensing image, the vegetation
Object includes landscape ground, crops, and the crops include cotton, interference crops;In classification layer L1In, build multidate light
Spectrum signature grader f (L12), landscape ground all in remote sensing image is rejected, remaining vegetation terrestrial object information is agriculture in remote sensing image
Crop, i.e. cotton, interference crops.
4.2.2 the remote sensing image of the crucial phase of first gradient cotton identification, is chosen, the 2nd layer of structure classification layer is denoted as L2,
In classification layer L2In, establish multidate spectral signature grader f (L2), the interference crops of spring riotous growth are rejected, with Hebei
For province Nangong City, the interference crops winter wheat of spring riotous growth, remaining vegetation terrestrial object information includes in remote sensing image
Cotton, corn and soybean.
4.2.3 the remote sensing image of the crucial phase of the second gradient cotton identification, is chosen, the 3rd layer of structure classification layer is denoted as L3,
In classification layer L3In, establish multidate spectral signature grader f (L3), reject the interference crops corn, big that bare area feature is presented
Beans, remaining vegetation terrestrial object information is cotton and peanut in remote sensing image.
4.2.4 the remote sensing image of the crucial phase of 3rd gradient cotton identification, the 4th layer of L of structure classification layer, are chosen4, dividing
Class layer L4In, establish multidate spectral signature grader f (L4), reject the interference crops peanut that bare area feature is presented, remote sensing shadow
Remaining vegetation terrestrial object information is the preliminary information result of cotton as in.
In formula, VndviExponential quantity, V are normalized for vegetationBrightnessFor wave band average brightness value, VGLCMContrast(all dir)For
Texture comparison's angle value, VBlue_MeanFor blue wave band wave band average value, VMax.diffFor maximum heterogeneity index value, a, b, c, d, e, f,
G, h is threshold value.
Step 5: preliminary information result to cotton in the remote sensing image that is obtained in step 3 and may mix and be divided into cotton
Other interference secondary multi-scale divisions of crop information, build grader, finely extract Cotton information.
5.1, the preliminary information result to cotton in the remote sensing image that is obtained in step 4 and its for being divided into cotton may be mixed
He interferes crop information to carry out secondary multi-scale division.
It is peanut, soybean with mixed point of interference crops of cotton;The remote sensing shadow that the secondary multi-scale division is based on
As being the remote sensing image corresponding to the crucial phase of the last one gradient cotton identification.
Secondary multi-scale division choice of optimal scale:Secondary multi-scale division object is as target area, to interfere farming
The minimum remote sensing image near infrared band mean variance of object is assessment parameter, is segmentation scale lower limit with 3, and 2 be step-length, in step 3
The optimum segmentation scale of initial partitioning is the upper limit, and it is secondary multiple dimensioned point to select corresponding segmentation scale when mean variance value maximum
The optimum segmentation scale cut.
5.2, grader is built.In conjunction with cotton spectral signature curve and cotton textural characteristics structure grader f (L), such as
Shown in following formula, rejected from the cotton preliminary information result of step 4 and the mixed interference crops divided of cotton, extraction cotton subclass 1
Information, 3 information of 2 information of cotton subclass and cotton subclass that extraction leakage divides respectively from other interference crops peanut, soybean.
Wherein, VndviExponential quantity, V are normalized for vegetationBrightnessFor wave band average brightness value, VGLCM Contrast(all dir)
For texture comparison's angle value, VMean4For the wave band average value of near infrared band, k, l, m, n, o, p are threshold value.
Step 6: carrying out merger, output drawing to the cotton various information of extraction.
Below by taking Hebei province Nangong City as an example, further details of theory is made to the present invention in conjunction with the drawings and specific embodiments
It is bright.
When a kind of cotton remote-sensing monitoring method based on phenology analysis of the present invention shoots more suitable for same satellite
Phase remote sensing image, specific implementation flow as shown in Figure 1, include the following steps:
Step 1: calmodulin binding domain CaM cotton Spectral Characteristics Analysis and the analysis of chief crop phenology, specify cotton remote sensing monitoring
Crucial phase;
1.1, cotton planting phase main interference crop type and its phenology are determined, phenology table is made;
Nangong City interference crop type mainly has winter wheat, corn, peanut and soybean, and phenology analytical table is as shown in Figure 3.
1.2, in conjunction with phenology table, cotton spectral signature curve is drawn, and signature analysis is carried out to cotton spectral signature;
In the present embodiment, with having cotton sample totally 292 pieces of Nangong City, as shown in fig. 4 a, are adopted based on Neuman (2011)
Then, oversampling ratio 0.3 selects 88 pieces of cotton plot 292*0.3 ≈ to control gauge, as shown in Figure 4 b, cotton after having and selecting
Plot distribution is as shown in Figure 4;According to Cotton Phenology analytical table, cotton sample prescription different growing stages phase, respectively April 21 are selected
Day, May 16, June 22, August 27 days, September 8 days, 6 phase high score No.1 multispectral images on October 14, calculate each cotton
The average value of 4 wave band DN values of area image corresponding to sample prescription:Sample prescription DNs all to each wave band of each growth period
Average value takes mean value again, obtain cotton sample prescription each wave band of each growth period value,It is difference to draw abscissa
Wave band, ordinate are6 cotton indicatrixes, as shown in Figure 5;Drafting abscissa is different phases, and ordinate is4 cotton indicatrixes;As shown in Figure 6.
Sowing time April 21 and seeding stage spectral signature on May 16 meet bare area spectral signature;June 22, flower bud phase belonged to
Vegetation growth initial stage meets basic vegetation spectral features;August flowering and boll-setting period on the 27th and the transition stage in the term of opening bolls and 9
The term of opening bolls on the moon 8 belongs to the vegetation growth vigorous stage, has typical vegetation spectral signature;The ripe harvest time cotton on October 14
The flowers are in blossom begins to wither, and vegetation characteristics gradually weaken, but have basic vegetation spectral features;
1.3, in conjunction with the curve of spectrum signature analysis of cotton different stages of growth, and it is based on remote sensing image visual interpretation, built
The interpretation mark in vertical cotton growth stage, Fig. 7 give cotton growing stage interpretation mark schematic diagram, and (a) April 21, cotton was in
In sowing time, (b) May 16, cotton was in the seeding stage, and bare area information is presented in image;(c) cotton on June 22 is flower bud phase, is in
The primary stage of growth, vegetation can cover earth's surface, pinkiness, and the terrain surface specifications at cotton vegetation initial stage are presented in image;(d) August
Cotton on the 27th is the transition stage of flowering and boll-setting period and the term of opening bolls, and growth is vigorous, is in rose, image is with being presented the luxuriant vegetation of cotton
Region feature;(e) October 14 days, cotton is harvest time, but leaf etc. is dirty-green, is in light red on image, it is withered that cotton is presented
Vegetation terrain surface specifications, match with spectrum analysis.
1.4, gradient is divided chooses with the crucial phase of cotton identification in gradient;
First gradient:It does not sow, sowing time and seedling stage, phase screen agrotype needed to be considered and is:Cotton, winter are small
Wheat, corn, peanut, soybean.The gradient phase coverage area is late May mid-April-, and opposite with cotton topographical features is dry
It disturbs crop specie number and is up to 1, the crucial phase value of cotton identification is respectively in 4 months, the last ten-days period and upper, middle and lower in May
Ten days, as shown in Figure 8 a, corresponding when this is several and cotton performance characteristic winter wheat, therefore rejecting winter wheat by contrast, the
Two gradient agrotypes are corn, cotton, soybean, peanut.
Second gradient:Flower bud phase, flowering and boll-setting period, the term of opening bolls, which is late September early June-, with cotton
The opposite interference crop specie number of flower topographical features is up to 2, and the crucial phase value of cotton identification is in 6 months, the last ten-days period, such as
Shown in Fig. 8 b.Corresponding at this time and cotton performance characteristic corn and soybean, therefore rejecting corn and soybean by contrast, third
Gradient agrotype is cotton, peanut.
3rd gradient:Ripe harvest time, which is early November early October-, with cotton earth's surface
The opposite interference crop specie number of feature is up to 1, and the crucial phase value of cotton identification is respectively upper, middle and lower ten days in October
And early November, as shown in Figure 8 c.Corresponding at this time and cotton performance characteristic peanut, therefore rejecting by contrast, the 4th ladder
Degree agrotype is cotton.
Therefore, it is analyzed in conjunction with cotton Spectral Characteristics Analysis and region phenology, and combines the availability of remotely-sensed data, it is final true
Fixed cotton remote sensing monitoring key phase is April 21, June 22, October 14.
Step 2: flowering and boll-setting period residing for cotton and choosing a phase vegetation growth vigorous period between the phase in the term of opening bolls
Remote sensing image.In conjunction with the availability of remotely-sensed data, increased phase remote sensing monitoring key phase is August 27.
Step 3: obtaining crucial phase remote sensing image, and pre-processed;
The Nangong City high score No.1 mostly light in April 21, June 22, August 27 days and October 14 is obtained in the present embodiment
Image is composed, image data is pre-processed, multispectral image wave band is four near-infrared, red, green, blue wave bands, to image profit
It is to carry out geometric exact correction with reference to image with Landsat/TM30 meters of orthographies, correction precision is the picture of region of no relief 0.5-1
Member;Then it is reference with image on April 21, image registration is carried out to other 3 phase images, ensures that 4 phase images do not misplace.
Step 4: carrying out multi-scale division to multidate image, layering structure grader tentatively extracts cotton;
Multi-scale division is carried out to multidate image, optimum segmentation scale is while meeting each issue of image object spectrum segmentation
Index be peak value corresponding to each phase divide scale-value average value, be computed, 4 phase of Nangong City image multi-scale division it is optimal
It is 15 to divide scale.L1For August image on the 27th, L2For image on April 21, L3For image on June 22, L4For image on October 14.
Every layer of grader is specific as follows:
Wherein, VndviExponential quantity, V are normalized for vegetationBrightnessFor wave band average brightness value, VGLCMContrast(all dir)For
Texture comparison's angle value, VBlue_MeanFor blue wave band wave band average value, VMax.diffFor maximum heterogeneity index value.
Step 5: to cotton PRELIMINARY RESULTS and a point object secondary splitting may be mixed, in conjunction with cotton spectral signature, textural characteristics
Grader is built, mixed point of removal, supplement leakage carries Cotton information;
(1) secondary splitting object select.The cotton of secondary splitting object select Nangong City tentatively extracts result and Nangong City
It is easier to cotton mixed point of interference crop peanut and soybean in other crops, amounts to 3 classes, when image data selection and withdrawal cotton is dry
The image for disturbing crop minimum period has chosen 16 meters of images of high score No.1 on October 14.
(2) secondary splitting choice of optimal scale.Using the cotton of selection tentatively extracted, peanut, soybean as target, with October
Image near infrared band mean variance on the 14th is secondary splitting scale assessment parameter, is segmentation scale lower limit with 3,2 be step-length, just
Secondary segmentation optimal scale 15 is the upper limit, and the optimal scale of secondary splitting is 5.
(3) grader is built.In conjunction with the cotton spectral signature curve of Nangong City, as shown in fig. 6, and cotton textural characteristics
The grader that structure cotton is finely extracted, extracts cotton subclass 1, Extraction of Peanut goes out levant cotton seed from the cotton tentatively extracted
Class 2, Extraction of Peanut go out cotton subclass 3.
Wherein, VndviExponential quantity, V are normalized for vegetationBrightnessFor wave band average brightness value, VGLCM Contrast(all dir)
For texture comparison's angle value, VMean4For the wave band average value of near infrared band.
Step 6: carry out merger all kinds of to the Cotton information of Multi-layer technology and output are charted
Drawing utilizes ArcGIS softwares, determines engineer's scale, increases legend, compass, completes cotton distribution drawing, such as Fig. 9
It is shown.
In order to the method for the present invention effectively verify and evaluate, cotton sample prescription is utilized to the Nangong City Cotton information of extraction
Data have carried out precision test.
Steps are as follows for the attribute accuracy calculating of cotton remote sensing monitoring:First, by verification with sample prescription data by vector (.shp lattice
Formula) raster data is converted to, spatial resolution is identical as staple crops spatial distribution data;Then by grid sample prescription data
Switch to point data, the position of point is the center of each pixel;Then, point data and the Cotton information of extraction is vector superposed
Analysis, calculates the attribute accuracy of cotton remote sensing monitoring, and calculation formula is as follows:
In formula, SX_accuracy is attribute accuracy, NoverlayThe point intersected with cotton remote sensing monitoring vector for cotton sample prescription
Quantity, NsampleQuantity a little is changed into for cotton sample prescription.It is computed, Nangong City sample prescription data change into a little totally 4448, all fall
Point on cotton remote sensing monitoring vector is 4194, therefore, attribute accuracy 94.3%.
The area precision of cotton remote sensing monitoring calculates as follows:Verification is carried out with sample prescription vector and cotton remote sensing monitoring vector
Superposition, the area of both statistics intersection, the ratio of the area and cotton sample gross area is area precision, and calculation formula is such as
Under:
In formula, MJ_accuracy is area precision, AoverlapIntersect with cotton remote sensing monitoring vector for cotton sample prescription vector
The area in region, AsampleIt for the area of cotton sample prescription vector, is computed, Nangong City cotton sample prescription vector and cotton remote sensing monitoring
The area of vector intersection is 1078165.5m2, Nangong City cotton sample gross area is 1138033.8m2, therefore area precision is
94.7%.
Attribute accuracy is 94.3%, area precision is 94.7% requirement for reaching practical application.
Embodiments of the present invention are explained in detail above in conjunction with attached drawing, but the present invention is not limited to above-mentioned implementations
Mode, one skilled in the relevant art within the scope of knowledge, can also make many variations to it.