CN107590816A - A kind of Water-Body Information approximating method based on remote sensing images - Google Patents
A kind of Water-Body Information approximating method based on remote sensing images Download PDFInfo
- Publication number
- CN107590816A CN107590816A CN201710806824.1A CN201710806824A CN107590816A CN 107590816 A CN107590816 A CN 107590816A CN 201710806824 A CN201710806824 A CN 201710806824A CN 107590816 A CN107590816 A CN 107590816A
- Authority
- CN
- China
- Prior art keywords
- water
- remote sensing
- sensing images
- image
- body information
- 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
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The present invention relates to water body information and approximating method, more particularly to a kind of Water-Body Information approximating method based on remote sensing images, the present invention is difficult timely grasp water body change and variation of water to solve existing general measure, and general measure is possible to find the shortcomings that feature of some pollution sources and pollution sources, and a kind of Water-Body Information approximating method based on remote sensing images is proposed, including：Remote sensing images are handled using water body index method, the image after being handled；Two-dimentional Otsu Threshold segmentations are used to the image after processing, obtain Threshold segmentation result；Threshold segmentation result and the image after processing are taken into common factor, it is determined that taking the light reflectivity of the image after occuring simultaneously；Selection of Function model, is respectively calculated using light reflectivity and function model to dissolved oxygen amount and permanganate, obtains the fitting degree of dissolved oxygen amount and the fitting degree of permanganate；Choose that fitting degree is optimal to be fitted respectively according to default standard.The present invention is applied to Water-Body Information and is fitted.
Description
Technical field
The present invention relates to water body information and approximating method, and in particular to a kind of Water-Body Information based on remote sensing images is intended
Conjunction method.
Background technology
The method of conventional water analysis is to sample on the spot, i.e., by long lasting for monitoring and record pollutant in water body
Species and concentration, according to recorded data, analyze water pollution variation tendency, and grading evaluation water quality condition.In order to observe not
Sampled when needing with time different zones and carry out and sample more more, in order to have deeper to observing regional water pollution variation tendency
Solution, it is necessary to after sampling data point handled accordingly, such as rejecting abnormalities point, expend huge manpower, material resources and financial resources；
With population increases, city enlarging, industrial or agricultural pollution etc., water pollution degree constantly deteriorates, and the diffusion of water pollution regional sustained, carries out
Part water body sample after sampling can not represent the distribution characteristics and variation of water of all areas water pollution；The earth's surface water capacity
It is vulnerable to the influence of the human factors such as climate change, land cover pattern, and the spy such as highly dynamic and uneven distribution of water body
Property cause by general measure to be difficult timely to grasp water body change and variation of water；With the gradual increasing of pollution source category
It is more, it is possible to find the feature of some pollution sources and pollution sources by general measure.
The content of the invention
It is difficult timely grasp water body change and change of water quality feelings the invention aims to solve existing general measure
Condition, and the shortcomings that general measure is possible to that some pollution sources and pollution sources feature can not be found, and propose one kind and be based on remote sensing figure
The Water-Body Information approximating method of picture.
A kind of Water-Body Information approximating method based on remote sensing images, including：
Step 1: remote sensing images are handled using water body index method, the image after being handled；
Step 2: using two-dimentional Otsu Threshold segmentations to the image after processing, Threshold segmentation result is obtained；
Step 3: the image after the Threshold segmentation result and the processing is taken into common factor, it is determined that taking the image after occuring simultaneously
Light reflectivity；
Step 4: the Selection of Function model successively from default collection of functions, uses light reflectivity and the function model
Dissolved oxygen amount and permanganate are respectively calculated, obtain the fitting degree and height of dissolved oxygen amount corresponding to each function model
The fitting degree of manganate；
Step 5: according to default standard scores from the fitting degree of the dissolved oxygen amount and the fitting degree of permanganate
Xuan Qu fitting degree be optimal is fitted.
Beneficial effects of the present invention are：1st, water body change and variation of water can be grasped in time；2nd, it is dirty to be more easy to extraction
Dye source and the feature of pollution sources；3rd, cost of the present invention is low, it is easy to accomplish.
Brief description of the drawings
Fig. 1 is the flow chart of the Water-Body Information approximating method based on remote sensing images of the present invention；
Fig. 2 is division of the Threshold segmentation vector of the present invention to two-dimensional histogram；
Fig. 3 normalizes difference water body index method, the extraction comparative result figure of two-dimentional Otsu thresholding methods to be improved, its
Middle Fig. 3 a are image after RGB synthesis, and Fig. 3 b are the image after improved normalization difference water body index method processing, and Fig. 3 c are two dimension
Image after the processing of Otsu Threshold segmentations；
Fig. 4 is permanganate matched curve figure in embodiment five；
Fig. 5 is the curve map of dissolved oxygen amount inverse model in embodiment five；
Fig. 6 is the remote sensing images after registration in embodiment seven, wherein Fig. 6 a be Song Hua River in May, 2014 with
In May, 2015 registration image；Fig. 6 b are the waters in May, 2015 image registering with June, 2015；Fig. 6 c are the waters 2015
May image registering with July, 2015；Fig. 6 d are the waters in May, 2015 image registering with October, 2015；
Fig. 7 is the region of variation figure in the waters in Fig. 6, and wherein Fig. 7 a are with respect to 2014 in May, 015 of 5 monthly variation figures, are schemed
7b is in June, 2015 relative to 5 monthly variation figures in 2015；Fig. 7 c are with respect to 2015 in July, 2015 of 5 monthly variation figures, and Fig. 7 d are
In October, 2015 was relative to 5 monthly variation figures in 2015；
Fig. 8 is the inversion result figure of 2016, and wherein Fig. 8 a are the inversion result figure of in September, 2016, and Fig. 8 b are 2016
The inversion result figure in October；
Fig. 9 is the inversion result figure of 2015, and wherein Fig. 9 a are the inversion result figure in April, 2015, and Fig. 9 b are 2015 7
The inversion result figure of the moon, Fig. 9 c are the inversion result figure in October, 2015；
Figure 10 is Song Hua River waters permanganate content distribution figure；
Figure 11 is Song Hua River waters dissolved oxygen amount content distribution figure.
Embodiment
Embodiment one：The Water-Body Information approximating method based on remote sensing images of present embodiment, including：
Step 1: remote sensing images are handled using water body index method, the image after being handled.
Step 2: using two-dimentional Otsu Threshold segmentations to the image after processing, Threshold segmentation result is obtained.
Specifically, Water-Body Information is obtained from different remote sensing images, the first of research typically related to water body
Step.Traditional Clean water withdraw method belongs to from principle extracts water body based on spectral characteristic, that is, utilizes water body and other ground species
Not obvious spectral characteristic (water body is high near infrared band absorptivity, low in visible light wave range reflectivity), and the present invention uses
Water body index method is handled multi-spectral remote sensing image, is strengthened water body and is suppressed non-water body, chooses manually on this basis
Suitable threshold value carries out Clean water withdraw.
The present invention carries out water body using two-dimentional Otsu Threshold segmentations on the basis of normalization difference water body index is improved and carried
Taking, the method is to choose appropriate threshold automatically according to the two-dimensional histogram of image, and applicability is stronger compared with manual selected threshold,
Precision is higher.
Step 3: the image after the Threshold segmentation result and the processing is taken into common factor, it is determined that taking the image after occuring simultaneously
Light reflectivity.
Step 4: the Selection of Function model successively from default collection of functions, uses light reflectivity and the function model
Dissolved oxygen amount and permanganate are respectively calculated, obtain the fitting degree and height of dissolved oxygen amount corresponding to each function model
The fitting degree of manganate.
Water quality monitoring be typically according to spectral reflectivity and actual measurement water quality parameter concentration between correlation, choose one or
Several wave bands carry out linear or nonlinear regression, draw regression model as dependent variable.Carry out the data of water quality monitoring needs
Generally there are the spectral reflectivity and sample point data on the spot of remote sensing images.Wherein, remotely-sensed data preferably with sampled data on the spot
Time be consistent, it is ageing to ensure；Secondly, sample point data is unsuitable very few, to ensure the accuracy of model and fit
The property used.Ammonia-nitrogen content, oxygen content in water and permanganate content and single band reflectivity or band combination reflectivity have certain
Correlation.It is and bigger in their sensitive band, coefficient correlation.
This thought is based in the present invention, using nearly 4 years remote sensing images of Landsat 8 and Heilongjiang Province Zhaoyuan website (east
Through 124 ° 59 ' 20 ", the measurement data of north latitude 45 ° 28 ' 15 "), spectral reflectivity and the correlation of measured data are analyzed, is found out quick
Feel wave band, and analyze under exponential function, power function and polynomial function form, the inverting of permanganate and dissolved oxygen in water
Models fitting situation, and error sum of squares (Sum of the Squared Errors) according to fitting, determine property coefficient (R-
Square), root-mean-square error (Root Mean Squared Error), the revised coefficient of determination (Adjusted R-
Square suitable inverse model) is selected.It is unsuccessful that Adjusted R-square are that negative is then fitted, should be closer to 1 better；
R-square sections are [- 1,1], and absolute value is better closer to 1；SSE and RMSE are better closer to 0.
Step 5: according to default standard scores from the fitting degree of the dissolved oxygen amount and the fitting degree of permanganate
Xuan Qu fitting degree be optimal is fitted.
The present invention carries out water quality monitoring using remote sensing technology, with monitored area is big, speed is fast, is easy to long term monitoring etc. excellent
Point, more accurately to grasp water body and water quality comparison, the present invention applies on improved normalization difference water body index
Two-dimentional Otsu Threshold segmentations carry out Clean water withdraw, and according to remote-sensing image spectrum reflectivity on the basis of Clean water withdraw and on the spot
Correlation between reference data establishes the inverse model of water quality parameter, and Border in Harbin Area water body is carried out by the model of foundation
Water analysis, for the method compared to more conventional water analysis, the present invention has cost low, it is easy to accomplish the advantages that.
Embodiment two：Present embodiment is unlike embodiment one：In step 1, the water body refers to
The computational methods of normalization difference water body index are in number method：
MNDWI=(Green-SWIR1)/(Green+SWIR1)
Wherein MNDWI is normalization difference water body index, and SWIR1 is short infrared wave band.
Specifically, in Different Waters extracting method, most widely used is water body index method.It is to pass through enhancing
Difference between water body and adjacent pixel protrudes water body interested, while suppresses most of noise.
On the basis of improved normalization difference water body index (NDWI), Xu Hanqiu discoveries are replaced with short infrared wave band
It near infrared band, can more strengthen the contrast between water body and background than NDWI, and propose improved normalization difference
Water body index (MNDWI)：
MNDWI=(Green-SWIR1)/(Green+SWIR1) (0-1)
Other steps and parameter are identical with embodiment one.
Embodiment three：Present embodiment is unlike embodiment one or two：In step 2, Otsu thresholds
Value segmentation detailed process be：
According to formulaBy f (x, y)≤s* and g (x, y) ＞ t*；And f (x, y) ＞
S* and g (x, y)≤t* pixel are ignored, and are arranged to 0 or 1.
Two-dimentional Otsu thresholding methods are to divide two-dimensional histogram region using gradient grayscales, and project Nogatas at two
On figure using traditional Otsu Threshold segmentations twice, it is therefore an objective to separate area-of-interest and background.This method has to noise
Stronger robustness, and arithmetic speed is very fast.
Assuming that a width gray level image f (x, y) (1≤x≤M, 1≤y≤N) size is M × N, each picture in the picture
Vegetarian refreshments f (x, y) calculates the average gray value of n × m neighborhoods, can obtain the smooth g (x, y) of a width, both tonal gradations are all
For 0,1 ... L.Wherein n × m neighborhood can be characterized by following template：
L typically takes 255, n and m to take the number more than 1, and they can be with unequal.The selection of neighborhood in currently associated document
There are various template, the neighborhood template of generally use 8 or 4 neighborhood templates.8 neighborhood templates are chosen in the present invention.
For each pixel in image, we can obtain a pair (i, j), and wherein i occurs from the original in f (x, y)
Beginning gray level, j are the neighborhood averaging gray levels occurred in g (x, y).Make cijThe probability of (i, j) is represented, its joint probability isWhereinJ=0,1 ..., L-1.
Given any threshold is vectorial (s, t).A and D represent object and background respectively in region.Region B and C represent edge and made an uproar
Sound.Allow two class C0And C1Object and background are represented respectively；ωiAnd μiClass C is represented respectivelyiProbability and average value vector.
C0And C1Probability can be expressed as：
C0And C1Average value vector can be expressed as：
The overall average vector of 2D histograms is：
In most cases, can be ignored away from cornerwise probability.It is easy to draw：ω0+ω1≈ 1, and
μT≈ω0μ0+ω1μ1。
Discrete matrix is defined as between class：
The trace description of discrete matrix is：
Therefore, optimal threshold can be expressed as：
Meet condition：f(x,y)≤s*And g (x, y) ＞ t*；And f (x, y) ＞ s*And g (x, y)≤t*Pixel neglected
Slightly, and it is arranged to 0 or 1.Fig. 2 shows division of the Threshold segmentation vector to two-dimensional histogram.
The water body information result of present embodiment is analyzed below by experiment：
Whether accurate in order to compare extraction result, it is a kind of relatively good method to carry out colored synthesis, is easy to contrast.This hair
It is bright to be tested using the 1T level products of Landsat 8 obtained from USGS.In order to strengthen water body, using Green, Red and
Tri- wave band weighting processing of NIR regenerate green band.That is R：Red, G：(Green+Red+NIR)/3, B：Green, obtain
New wave band, then three wave bands are synthesized, RGB shows that composite result is as shown in Figure 3 a.In the remote sensing images of Landsat 8,
Green, Red, NIR correspond to respectively third and fourth, five wave bands.
Improved normalization difference water body index method, the extraction result of two-dimentional Otsu thresholding methods are respectively such as Fig. 3 b, figure
Shown in 3c.
In order to more intuitively compare two kinds of extraction accuracy, water body in the extraction result of every kind of method and true value figure is believed
Breath is compared, and according to correct recall rate P (TA), false alarm rate P (FA) and loss P (MA) definition：
Calculate correct recall rate, false alarm rate and loss, analysis precision, as shown in table 1-1.
Table 1-1 extraction accuracies
Other steps and parameter are identical with embodiment one or two.
Embodiment four：Unlike one of present embodiment and embodiment one to three：In step 4, in advance
If collection of functions include：First order exponential function, bi-exponential function, a rank multinomial, second order polynomial, three rank multinomials, one
Rank power function and second order power function.
In general, there are two methods can be with founding mathematical models：The first is mechanism based method analysis, and another kind of is test
Analysis method.The model established by mechanism based method analysis generally has clear and definite meaning.Test analysis be can not Direct Analysis grind
In the case of studying carefully object internal mechanism, research object is considered as a black box, measures the input and output by black box
Data, and the method for using statistical analysis on this basis, select one and measurement in certain model I determined
Best one model of data fitting effect.
Building for inverse model immediately belongs to method for testing and analyzing in the present invention.Conventional inverse model function shape is provided first
Formula, secondly, the correlation between spectral reflectivity and reference data concentration is analyzed, choose one or two most strong ripple of correlation
Duan Jinhang regression analyses, the best model of fitting effect is chosen according to mean square error of fitting etc..
In statistics, regression analysis refers to analyze the complementary relation i.e. correlation between two or more variables
A kind of statistical analysis technique.According to the quantity of variable, simple regression and multiple regression analysis can be divided into；According to independent variable and because
Relation between variable, linear regression analysis and nonlinear regression analysis can be divided into.
The water quality parameter inverse model functional form selected in the present invention is as shown in table 2-1.
Correlation is gone out according to spectral reflectivity and water quality parameter concentration analysis, select a stronger wave band of correlation or
Several wave bands carry out regression analysis, establish inverse model.Wherein, spectral reflectivity and reference data concentration such as table 2-2, table 2-
Shown in 3.
Table 2-2 spectral reflectivities
Table 2-3 reference data concentration
Other steps and parameter are identical with one of embodiment one to three.
Embodiment five：Unlike one of present embodiment and embodiment one to four：In step 4, intend
Conjunction degree specifically includes：Error sum of squares, determine property coefficient, root-mean-square error, the revised coefficient of determination.
Present embodiment on the basis of embodiment four, further by error sum of squares, determine property coefficient,
Square error, the revised coefficient of determination carry out the fitting degree of judgment curves, choose the optimal function model conduct of fitting degree
Final selection.
The detailed process of one embodiment is：
By reflectivity and reference data by correlation statistics, find the high violent hydrochlorate index of inverting and spectral reflectivity it
Between correlation degree.Coefficient correlation is as shown in Table 2-4.
Coefficient correlation between table 2-4 reflectivity and reference data
As can be seen that preceding several wave bands and permanganate correlation are stronger, respectively with second band reflectivity, the 3rd wave band
Reflectivity, the wave band reflectivity product of second band the 3rd, the wave band reflectivity of second band the 3rd and as independent variable, corresponding height
Mangaic acid salt index content is as dependent variable, with functional forms such as power function, exponential function, a rank multinomial, the multinomial forms of second order
It is fitted.Over-fitting can be caused because functional form exponent number is too high, function change is violent, does not meet actual water quality parameter change
Trend；Functional form exponent number is too low to cause some point fittings unsuccessful, precision reduction.Therefore, it is (ρ from independent variable2+ρ3),
That is the sum of second, third wave band reflectivity, dependent variable are permanganate content, and functional form is bi-exponential form, fitting effect
Fruit is preferable.Wherein, the matched curve of permanganate is as shown in Figure 4.
Functional form：Y=aebx+c·edx
Parameter value：A=13.17；B=-8.742；C=1.321；D=4.726
Fitting degree：SSE=0.1098；R-square=0.9596；Adjusted R-square=0.9191；
RMSE=0.1914
By reflectivity and reference data by correlation statistics, find and associate journey between dissolved oxygen amount and spectral reflectivity
Degree.Coefficient correlation is as shown in table 2-5.
Coefficient correlation between table 2-5 reflectivity and reference data
As can be seen that the correlation between dissolved oxygen amount and each wave band is not strong, most it is just by force with first band correlation
0.5942.Multiple band spectrum reflectivity can be combined according to certain mathematical relationship, then complementary operation is carried out with dissolved oxygen amount, can
It can be increased considerably by property.Respectively with first band reflectivity, second band reflectivity, the first second band reflectivity sum,
First second band reflectivity product is as independent variable, and corresponding dissolved oxygen amount content is as dependent variable, with power function, index letter
The functional forms such as number, a rank multinomial, the multinomial form of second order are fitted.Test result indicates that reflected with first, second wave band
For the product of rate as independent variable, dissolved oxygen amount content is dependent variable, and type function is best for the fitting degree of three rank multinomials.Its
In, the matched curve of dissolved oxygen amount is as shown in Figure 5.
Functional form：Y=p1·x3+p2·x2+p3·x+p4
Parameter value：p1=1195000；p2=-91810；p3=2183；p4=-8.265
Fitting degree：SSE=5.294；R-square=0.7923；Adjusted R-square=0.5846；RMSE=
1.328
Other steps and parameter are identical with one of embodiment one to four.
Embodiment six：Unlike one of present embodiment and embodiment one to five：
Step 5 one, four kinds of evaluation indexes are established, be specially：
The revised coefficient of determination is closer to 1, then fitting degree is better；
The absolute value of property coefficient is determined closer to 1, then fitting degree is better；
Error sum of squares error is closer to 0, then fitting degree is better；
Root-mean-square error is closer to 0, then fitting degree is better；
Step 5 two, for dissolved oxygen amount and permanganate content, after calculating the amendment corresponding to each function model
The coefficient of determination, determine the absolute value of property coefficient, error sum of squares, root-mean-square error；
Step 5 three, count that each function model how many evaluation index is higher than other function models；And select symbol
The most function model of the evaluation index quantity of conjunction condition.
I.e. present embodiment judges whether fitting degree is good using 4 indexs, and selection meets index and does more function progress
Fitting.If the eligible index quantity of multiple functions is all most simultaneously, just all chooses, be all fitted, Ran Houcong
It is fitted the conduct final choice chosen in the image drawn and best suit actual landform situation.
Present embodiment provides a kind of method for choosing the optimal function model of fitting degree, in real process, is
Guarantee fitting result more tallies with the actual situation, and can be checked and approved by manually doing last verification, filter out and intuitively more meet water
The model of body objective condition.
Other steps and parameter are identical with one of embodiment one to five.
Embodiment seven：Unlike one of present embodiment and embodiment one to six：Held in step 1
Before row, in addition to Multitemporal Remote Sensing Images step of registration, specifically include：
Step A, Harris Corner Feature extractions are carried out to Multitemporal Remote Sensing Images；
Step B, reject error hiding using normalized crosscorrelation matching algorithm and mismatch vector, obtain the figure after registration
Picture；
Step C, the processing of step 1 is carried out using the image after registration as remote sensing images.
Present embodiment is the water analysis for Multitemporal Remote Sensing Images, the content of present embodiment be in order in order to
Carry out rocking bar image registration, registration after image can be a kind of as step image input.
Specifically, in order to intuitively water outlet body region situation of change, it is necessary to become to Multitemporal Remote Sensing Images
Change detection, need first to carry out the registration between remote sensing images before this.Harris Corner Feature extractions are carried out first, are then utilized
Normalized crosscorrelation matching algorithm (NCC) is slightly matched, and is rejected error hiding and is mismatched vector, based on gray scale coefficient correlation,
Registration error is calculated so as to obtain the image after registration.
(1) Harris Corner Detection Algorithms
Harris proposes Harris Corner Detection Algorithms within 1988, if it is considered that the difference of weight is contributed in difference in window
Different, angle point receptance function can be written as：
Wherein,It is dimensional Gaussian window function.If it is considered that the diversity of offset direction, profit
It is approximate with single order Taylor, the angle point receptance function of Harris algorithms is can obtain by formula (4-1)：
Wherein, IxAnd IyIt is pixel (u, v) in first derivative both horizontally and vertically：
Three values are calculated pixel： The gradient matrix of pixel can then be drawn
Thus Harris angle point receptance function (CRF) expression formula obtains：
CRF (u, v)=det (M)-k (trace (M))2=(AB-C2)-k(A+B)2 (0-14)
Wherein, det (M)=AB-C2, trace (M)=A+B, k is constant, typically takes 0.04~0.06.As the CRF of target
When value is more than or equal to this threshold value, the pixel is judged as angle point.
(2) NCC matching algorithms
Normalized crosscorrelation (Normalized Cross Correlation method, NCC) matching algorithm is a kind of warp
The statistical match algorithm of allusion quotation, the degree of matching is determined by calculation template and the cross correlation value of matching.
According to the definition of vector dot：
Ab=| a | | b | cos θ (0-15)
If two vectors are similar, their direction is identical, and its angle is θ, therefore can be judged according to cos θ value
Two vectorial similitudes.It is generalized in two dimension, then
In formula, R (u, v) is the normalizated correlation coefficient of location point (u, v)；N1×N2For matching template size；xi+u,j+v,
yi,jRespectively need in two width that match (i+u, j+v), the gray value at (i, j) place.R (u, v) value is bigger, it was demonstrated that two width
Similitude is higher.
Result after registration is as shown in Figure 6.
Public domain is intercepted in registering every group, carries out Clean water withdraw, and calculate region of variation.Due to carrying out water body
Extraction carries out binaryzation equivalent to by gray scale, so, calculating region of variation is carried out using equation below：
Change=abs (imt1-imt2) (0-17)
Wherein, imt1 and imt2 is respectively the first width remote sensing images and the second width remote sensing images, and both, which make the difference, takes absolute value
Draw region of variation.Testing result is as shown in Figure 7.
By reference data on the spot it can be found that Song Hua River is classified according to dissolved oxygen amount, generally there was only I class and II class water body,
So for finer display analysis, finer classification is carried out for dissolved oxygen amount, is shown.
For IV class water body, it is divided into 3-5mg/L, for III class water body, is divided into 5-6mg/L, by III class water body and IV class water
The decorum one is shown with green；For II class water body, more meticulously it is divided into 6-6.5mg/L, 6.5-7mg/L, 7-7.5mg/L, respectively
Shown with blueness, yellow, pink colour；For I class water body, more meticulously it is divided into 7.5-10mg/L, more than 10mg/L, respectively with orange
Color, cyan are shown.Wherein, ammonia-nitrogen content is lower, and water quality is poorer, by I class to V class to bad V (dissolved oxygen amount concentration is less than 2mg/L)
Class, water quality is worse and worse.Exemplified by 2016, inversion result is as shown in Figure 8.
Inversion result is analyzed：Because permanganate concentration concentrates on the IVth class water in Songhua River Harbin water body
In the range of body, i.e. 6-10mg/L.After inverting, permanganate content is classified, with different colors to different permanganates
The region of content is shown.
By the unified green display of I class water body, II class water body and III class water body；For IV class water body, more meticulously it is divided into
6-6.5mg/L, 6.5-7mg/L, 7-7.5mg/L, 7.5-10mg/L, shown respectively with blueness, yellow, orange, pink colour；By V class
Water body and the unification of bad V class (permanganate concentration is more than 15mg/L) water body are represented with cyan.Wherein, permanganate content is lower,
Water quality is better, and by I class to V class to bad V class, water quality is worse and worse.Exemplified by 2015, inversion result is as shown in Figure 9.
Experimental result is analyzed：
Different Waters are counted, ratio and different oxygen shared by the Different Waters of different permanganate contents is calculated
Different Waters shared by ratio as shown in Figure 10, Figure 11,
Other steps and parameter are identical with one of embodiment one to six.
Analyze and draw from Figure 10, the basin permanganate content of Songhua River Harbin generally belongs to IV class water body, pole
Small part belongs to III class water body, and I class water body, II class water body, V class water body and bad V class water body are not present.IV class water body
In, permanganate content is largely belonged in the range of 6.5-7mg/L.Annual 9, or so October, the most of meeting of permanganate
Concentrate in the range of 6-6.5mg/L, water quality increases；When to summer, high violent phosphate content can rise to a 7mg/L left sides
The right side, water degradation.And the water quality inside the water-quality ratio water body at river course edge is worse.From permanganate angle analysis, Song Hua River
In the water pollution of Harbin basin, the pollution of permanganate is more serious, and water quality is poor.
Understand that the dissolved oxygen amount in SONGHUA RIVER OF HARBIN basin is concentrated in the range of 7.5-10mg/L, belongs to I according to Figure 11 analyses
Class water body.V class water body seldom is partly belonged to, i.e., concentration is in the range of 3-5mg/L.From oxygen content angle analysis, Song Hua River Ha Er
Pollution is not present in the water body in shore current domain, and water quality is preferable.
It can be predicted and drawn by Figure 10 and 11, permanganate content will be annual in the water body in SONGHUA RIVER OF HARBIN basin
6-10 months 7mg/L or so float, 6mg/L or so will be down in remaining month, but in general, water body is still fallen within
IV class water body, it should take measures to improve water quality；Annual October or November, dissolved oxygen amount in the water body of SONGHUA RIVER OF HARBIN basin
For extreme portions in the range of 7.5-10mg/L, water quality is preferable, and remaining month can float in the range of 7mg/L or so, but it is overall and
Speech, water body belongs to I class or II class water body, water quality are preferable.
The present invention can also have other various embodiments, in the case of without departing substantially from spirit of the invention and its essence, this area
Technical staff works as can make various corresponding changes and deformation according to the present invention, but these corresponding changes and deformation should all belong to
The protection domain of appended claims of the invention.
Claims (7)
- A kind of 1. Water-Body Information approximating method based on remote sensing images, it is characterised in that including：Step 1: remote sensing images are handled using water body index method, the image after being handled；Step 2: using two-dimentional Otsu Threshold segmentations to the image after processing, Threshold segmentation result is obtained；Step 3: the image after the Threshold segmentation result and the processing is taken into common factor, it is determined that taking the light of the image after occuring simultaneously Reflectivity；Step 4: the Selection of Function model successively from default collection of functions, using light reflectivity and the function model to molten Oxygen amount and permanganate are respectively calculated, and obtain the fitting degree and permanganic acid of dissolved oxygen amount corresponding to each function model The fitting degree of salt；Step 5: selected respectively according to default standard from the fitting degree of the dissolved oxygen amount and the fitting degree of permanganate The function model for taking fitting degree optimal is fitted.
- 2. the Water-Body Information approximating method according to claim 1 based on remote sensing images, it is characterised in that in step 1, The computational methods of normalization difference water body index are in the water body index method：MNDWI=(Green-SWIR1)/(Green+SWIR1)Wherein MNDWI is normalization difference water body index, and SWIR1 is short infrared wave band, and Green is the 3rd of remote sensing images Wave band.
- 3. the Water-Body Information approximating method according to claim 1 or 2 based on remote sensing images, it is characterised in that step 2 In, the detailed process of Otsu Threshold segmentations is：According to formulaBy f (x, y)≤s*And g (x, y) ＞ t*, and f (x, y) ＞ s*And g (x,y)≤t*Pixel ignore, and be arranged to 0 or 1；Wherein s*Represent the first threshold for Threshold segmentation, t*Represent to be used for threshold The Second Threshold of value segmentation, L are the total quantity of tonal gradation, and s, t are respectively to give two vectorial components of any threshold, f (x, Y) it is the pixel in image, wherein x is the abscissa of image slices vegetarian refreshments, and y is the ordinate of image slices vegetarian refreshments, and g (x, y) is pair Pixel f (x, y) calculates the value obtained after the average gray of n × m neighborhoods；Tr represents to seek the mark of matrix；Sb(s, t) is represented between class Discrete matrix.
- 4. the Water-Body Information approximating method according to claim 1 based on remote sensing images, it is characterised in that in step 4, Default collection of functions includes：First order exponential function, bi-exponential function, a rank multinomial, second order polynomial, three rank multinomials, Single order power function and second order power function.
- 5. the Water-Body Information approximating method according to claim 4 based on remote sensing images, it is characterised in that in step 4, Fitting degree specifically includes：Error sum of squares, determine property coefficient, root-mean-square error, the revised coefficient of determination.
- 6. the Water-Body Information approximating method according to claim 5 based on remote sensing images, it is characterised in that step 5 is specific For：Step 5 one, for dissolved oxygen amount and permanganate content, calculate corresponding to each function model it is revised certainly Determine coefficient, the absolute value for determining property coefficient, error sum of squares, root-mean-square error；Step 5 two, for each function model, if its revised coefficient of determination compared to other models closer to 1, Then count value adds 1；If for the absolute value of its decision property coefficient compared to other models closer to 1, count value adds 1；If its Error sum of squares is compared to other models closer to 0, then count value adds 1；If its root-mean-square error is compared to other models more Close to 0, then count value adds 1；Step 5 three, select the count value highest model function model optimal as fitting degree and be fitted.
- 7. the Water-Body Information approximating method according to claim 1 based on remote sensing images, it is characterised in that held in step 1 Before row, in addition to Multitemporal Remote Sensing Images step of registration, specifically include：Step A, Harris Corner Feature extractions are carried out to Multitemporal Remote Sensing Images；Step B, reject error hiding using normalized crosscorrelation matching algorithm and mismatch vector, obtain the image after registration；Step C, the processing of step 1 is carried out using the image after registration as remote sensing images.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710806824.1A CN107590816B (en) | 2017-09-08 | 2017-09-08 | Water body information fitting method based on remote sensing image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710806824.1A CN107590816B (en) | 2017-09-08 | 2017-09-08 | Water body information fitting method based on remote sensing image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107590816A true CN107590816A (en) | 2018-01-16 |
| CN107590816B CN107590816B (en) | 2021-06-15 |
Family
ID=61051032
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710806824.1A Active CN107590816B (en) | 2017-09-08 | 2017-09-08 | Water body information fitting method based on remote sensing image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107590816B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111426806A (en) * | 2020-03-27 | 2020-07-17 | 渤海大学 | Automatic monitoring and early warning method for freshness degree of aquatic product cold-chain logistics based on means of Internet of things |
| CN111949270A (en) * | 2020-07-27 | 2020-11-17 | 中国建设银行股份有限公司 | Method and device for sensing running environment change of process robot |
| JP2021032898A (en) * | 2019-08-22 | 2021-03-01 | ▲広▼州大学 | Water area identification method and device based on iterative classification |
| CN113326855A (en) * | 2021-06-22 | 2021-08-31 | 长光卫星技术有限公司 | Night lamplight remote sensing image feature extraction method based on two-dimensional Gaussian surface fitting |
| CN117649612A (en) * | 2023-10-19 | 2024-03-05 | 成都大学 | Satellite hyperspectral remote sensing data surface water body extraction method based on hybrid algorithm |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101950364A (en) * | 2010-08-30 | 2011-01-19 | 西安电子科技大学 | Remote sensing image change detection method based on neighbourhood similarity and threshold segmentation |
| CN105243367A (en) * | 2015-10-12 | 2016-01-13 | 水利部水利信息中心 | Method and device for monitoring scope of water body based on satellite remote sensing data |
| CN106525762A (en) * | 2016-11-07 | 2017-03-22 | 航天恒星科技有限公司 | Water quality monitoring method and water quality monitoring device based on adaptive model |
| US20170084017A1 (en) * | 2015-09-17 | 2017-03-23 | Industry Academic Cooperation Foundation Keimyung University | METHOD AND APPARATUS FOR PREDICTING CHLOROPHYLL-a CONCENTRATION IN RIVER USING SATELLITE IMAGE DATA AND NONLINEAR RANSAC METHOD |
-
2017
- 2017-09-08 CN CN201710806824.1A patent/CN107590816B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101950364A (en) * | 2010-08-30 | 2011-01-19 | 西安电子科技大学 | Remote sensing image change detection method based on neighbourhood similarity and threshold segmentation |
| US20170084017A1 (en) * | 2015-09-17 | 2017-03-23 | Industry Academic Cooperation Foundation Keimyung University | METHOD AND APPARATUS FOR PREDICTING CHLOROPHYLL-a CONCENTRATION IN RIVER USING SATELLITE IMAGE DATA AND NONLINEAR RANSAC METHOD |
| CN105243367A (en) * | 2015-10-12 | 2016-01-13 | 水利部水利信息中心 | Method and device for monitoring scope of water body based on satellite remote sensing data |
| CN106525762A (en) * | 2016-11-07 | 2017-03-22 | 航天恒星科技有限公司 | Water quality monitoring method and water quality monitoring device based on adaptive model |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2021032898A (en) * | 2019-08-22 | 2021-03-01 | ▲広▼州大学 | Water area identification method and device based on iterative classification |
| CN111426806A (en) * | 2020-03-27 | 2020-07-17 | 渤海大学 | Automatic monitoring and early warning method for freshness degree of aquatic product cold-chain logistics based on means of Internet of things |
| CN111949270A (en) * | 2020-07-27 | 2020-11-17 | 中国建设银行股份有限公司 | Method and device for sensing running environment change of process robot |
| CN113326855A (en) * | 2021-06-22 | 2021-08-31 | 长光卫星技术有限公司 | Night lamplight remote sensing image feature extraction method based on two-dimensional Gaussian surface fitting |
| CN113326855B (en) * | 2021-06-22 | 2022-07-12 | 长光卫星技术股份有限公司 | Night lamplight remote sensing image feature extraction method based on two-dimensional Gaussian surface fitting |
| CN117649612A (en) * | 2023-10-19 | 2024-03-05 | 成都大学 | Satellite hyperspectral remote sensing data surface water body extraction method based on hybrid algorithm |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107590816B (en) | 2021-06-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107590816A (en) | A kind of Water-Body Information approximating method based on remote sensing images | |
| Coburn et al. | A multiscale texture analysis procedure for improved forest stand classification | |
| CN104408707B (en) | Rapid digital imaging fuzzy identification and restored image quality assessment method | |
| CN110097034A (en) | A kind of identification and appraisal procedure of Intelligent human-face health degree | |
| CN110210448B (en) | Intelligent face skin aging degree identification and evaluation method | |
| CN110120056B (en) | Blood leukocyte segmentation method based on adaptive histogram threshold and contour detection | |
| CN110378232B (en) | Improved test room examinee position rapid detection method of SSD dual-network | |
| CN110033446A (en) | Enhancing image quality evaluating method based on twin network | |
| CN108052980A (en) | Air quality grade detection method based on image | |
| CN106203522B (en) | Hyperspectral image classification method based on three-dimensional non-local mean filtering | |
| CN103049893A (en) | Method and device for evaluating image fusion quality | |
| CN103198482B (en) | Based on the method for detecting change of remote sensing image that disparity map fuzzy membership merges | |
| CN108550145A (en) | A kind of SAR image method for evaluating quality and device | |
| CN115909256B (en) | Road disease detection method based on road visual image | |
| CN109829905A (en) | It is a kind of face beautification perceived quality without reference evaluation method | |
| CN111738997A (en) | Method for calculating new coronary pneumonia lesion area ratio based on deep learning | |
| CN103729462B (en) | A kind of pedestrian retrieval method blocked based on rarefaction representation process | |
| CN107610119A (en) | The accurate detection method of steel strip surface defect decomposed based on histogram | |
| CN105894520A (en) | Satellite image automatic cloud detection method based on Gaussian mixture model | |
| Wang et al. | The PAN and MS image fusion algorithm based on adaptive guided filtering and gradient information regulation | |
| CN104657714B (en) | Illumination symmetry merged with global illumination intensity without refer to human face light evaluation method | |
| CN110148126A (en) | Blood leucocyte dividing method based on color component combination and contour fitting | |
| CN105809092A (en) | Population target detection method and device thereof | |
| CN114299059A (en) | Method for judging scratch defects of unsorted casting blanks on surfaces of hot-rolled strip steel | |
| CN104050678A (en) | Underwater monitoring color image quality measurement method |
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 |