CN104359847A - Method and device for acquiring centroid set used for representing typical water category - Google Patents

Abstract

The embodiment of the invention discloses a method and device for acquiring a centroid set used for representing the typical water category. The method comprises the steps of acquiring water sample data; carrying out dimensionality reduction on reflectance spectral data of all water surface sampling points to obtain two principal component data of the reflectance spectra of each sampling point; determining the number of categories; carrying out fuzzy classification on the water surface sampling points for at least two times according to the two principal component data of the reflectance spectra of each sampling point, wherein all the times of fuzzy classification are carried out based on different distance measures; solving the intersection between the results obtained in the process of all the times of fuzzy classification and the initial centroid set corresponding to the ith category to obtain the centroid set corresponding to the ith category, wherein i is equal to 1, 2......, C. After the method and the device are adopted, the stability of the centroid set corresponding to the ith category is improved, and the centroid set has representativeness.

Description

Method and device for obtaining centroid sets representing typical water body categories

technical field

The present invention relates to the technical field of remote sensing, and more specifically, relates to a method and a device for obtaining a centroid set representing a typical water body category.

Background technique

Remote sensing inversion of water quality parameters in optically complex water bodies has always been a difficult point. At present, many studies use the strategy of classification inversion for remote sensing inversion, that is, first classify the remote sensing pixels, and then use different strategies for inversion for different types of remote sensing pixels to achieve the purpose of improving the inversion accuracy. The classification of remote sensing pixels can be divided into hard classification and fuzzy classification. Among them, fuzzy classification has the characteristics of improving the smoothness and stability of the inversion results, and is the main research direction in this field.

The fuzzy classification of remote sensing pixels is to calculate the weight coefficients of each pixel belonging to each category, and then classify the pixels according to the weight coefficients of each pixel belonging to each category, and obtain several centroid sets. Among them, the weight coefficient of each pixel belonging to each category is determined by the distance between the pixel and the centroid pixel of each category. When the distance between the pixel and the centroid pixel is smaller, it is considered that the pixel and the centroid image The better the similarity of the pixel, the greater the weight coefficient that the pixel belongs to the category represented by the centroid pixel. Therefore, how to obtain the most effective centroid set representing the typical water body category is one of the important research contents of using the fuzzy classification method to invert the water quality parameters of inland water bodies.

In the process of implementing the present invention, the inventors found that the centroid sets representing typical water body categories currently obtained through fuzzy classification have poor stability.

Contents of the invention

The purpose of the present invention is to provide a method and device for obtaining a centroid set representing a typical water body category, so as to improve the stability of the centroid set representing a typical water body category.

To achieve the above object, the present invention provides the following technical solutions:

A method for obtaining a set of centroids representative of typical water body classes, comprising:

Obtaining water body sample data, the water body sample data includes remote sensing reflectance spectral data of several water surface sampling points;

Through principal component analysis transformation, the reflectance spectrum data of several bands of each water surface sampling point is reduced in dimension, and two principal component data of the reflectance spectrum of each water surface sampling point are obtained;

Determine the number of categories C;

The fuzzy classification process is performed at least twice, and the distance measures based on each fuzzy classification process are different; the fuzzy classification process divides the several water surface sampling points into C types according to the two principal component data of the reflectance spectrum of each water surface sampling point , get C initial centroid sets;

Compute the intersection of the initial centroid sets corresponding to the i-th category obtained by each execution of the fuzzy classification process, and obtain the centroid set corresponding to the i-th category; where, i=1, 2, . . . , C.

In the above method, preferably, the determination of the number of categories C includes:

According to the two principal component data of the reflectance spectrum of each water surface sampling point, under different classification numbers, the BIC index of the described several water surface sampling points;

The number of categories corresponding to the minimum BIC index is determined as the number of categories C.

In the above method, preferably, the determination of the number of categories C includes:

According to two principal component data calculations of the reflectance spectrum of each water surface sampling point under different classification data, the Dunn'c index of described several water surface sampling points;

The number of categories corresponding to the maximum Dunn'c index is determined as the number of categories C.

In the above method, preferably, the water body sample data also includes: the water body environmental variables of the several water surface sampling points, and the water body environmental variables include: water quality parameters and inherent optical quantities of different components of the water body; After the centroid sets corresponding to categories, the method also includes:

Display the water body environmental variable parameters of each sample in the centroid set corresponding to the i-th category;

When receiving the elimination instruction triggered by the user, determine the target sample data according to the identification of the sample data carried in the elimination instruction, delete the target sample data, and re-execute the step of performing the fuzzy classification process at least twice.

In the above method, preferably, the sampling area of the water body sample data includes at least one of the following areas: inland waters and nearshore waters.

A means of obtaining a set of centroids representing a typical class of water bodies, comprising:

The sample acquisition module is used to acquire water body sample data, and the water body sample data includes remote sensing reflectance spectral data of several water surface sampling points;

A dimensionality reduction module is used to reduce the dimensionality of the reflectance spectrum data of several bands of each water surface sampling point through principal component analysis transformation, so as to obtain two principal component data of the reflectance spectrum of each water surface sampling point;

A determination module is used to determine the number of categories C;

The classification module is used to perform the fuzzy classification process at least twice, and the distance measures based on each fuzzy classification process are different; the fuzzy classification process divides the obtained several water surface data according to the two principal component data of the reflectance spectrum of each water surface sampling point The sampling points are divided into C categories, and C initial centroid sets are obtained;

The centroid set acquisition module is used to intersect the initial centroid set corresponding to the i-th category obtained by each execution of the fuzzy classification process, and obtain the centroid set corresponding to the i-th category; wherein, i=1, 2, ... ..., C.

In the above device, preferably, the determination module includes:

The first calculation unit is used to calculate the BIC index of the several water surface sampling points under different classification numbers according to the two principal component data of the reflectance spectrum of each water surface sampling point;

The first determining unit is configured to determine the category number corresponding to the minimum BIC index as the category number C.

In the above device, preferably, the determination module includes:

The second calculation unit is used to calculate the Dunn'c index of the several water surface sampling points under different classification data according to the two principal component data of the reflectance spectrum of each water surface sampling point;

The second determination unit is used to determine the number of categories corresponding to the maximum Dunn'c index as the number of categories C.

In the above-mentioned device, preferably, the water body sample data also includes: the water body environmental variables of the several water surface sampling points, and the water body environmental variables include: water quality parameters and inherent optical quantities of different components of the water body; the device also includes :

The display module is used to display the water body environment variable parameters of each sample in the centroid set corresponding to the i-th category;

A deletion module, configured to determine the target sample data according to the identification of the sample data carried in the removal instruction when receiving a removal instruction triggered by the user, delete the target sample data, and generate a trigger instruction to instruct the classification module The step of performing the fuzzy classification process at least twice is re-performed.

The above device, preferably, the sample acquisition module is specifically used to acquire water body sample data, the water body sample data includes remote sensing reflectance spectral data of several water surface sampling points; the sampling area of the water body sample data includes the following areas At least one of: inland waters and near-shore waters.

It can be seen from the above scheme that the present application provides a method and device for obtaining a centroid set representing a typical water body category, obtains water body sample data, and performs dimensionality reduction on the reflectance spectral data of each water surface sampling point to obtain the reflectance of each sampling point The two principal component data of the spectrum determine the number of categories, and at least twice classify the several water surface sampling points according to the two principal component data of the reflectance spectrum of each water surface sampling point; wherein, each fuzzy classification is based on The distance measure is different; the initial centroid set corresponding to the i-th category obtained by each fuzzy classification process is obtained, and the centroid set corresponding to the i-th category is obtained, where i=1, 2, ..., C . While improving the stability of the centroid set corresponding to the i-th category, it makes the centroid set more representative.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work.

Fig. 1 is the method for obtaining the centroid set representing a typical water body category provided by the embodiment of the present application;

Fig. 2 is a kind of implementation flowchart of determining the category number C provided by the embodiment of the present application;

FIG. 3 is another implementation flowchart for determining the number of categories C provided by the embodiment of the present application;

Fig. 4 is another implementation flowchart of the method for obtaining a centroid set representing a typical water body category provided by the embodiment of the present application;

FIG. 5 is a schematic structural diagram of a device for obtaining a centroid set representing a typical water body category provided by an embodiment of the present application;

FIG. 6 is a schematic structural diagram of a determination module provided in an embodiment of the present application;

FIG. 7 is another schematic structural diagram of the determination module provided by the embodiment of the present application;

FIG. 8 is another schematic structural diagram of a device for obtaining a centroid set representing a typical water body category provided by an embodiment of the present application.

The terms "first", "second", "third", "fourth", etc., if any, in the description and claims and the above drawings are used to distinguish similar parts and not necessarily to describe specific sequence or sequence. It is to be understood that the data so used are interchangeable under appropriate circumstances such that the embodiments of the application described herein can be practiced in sequences other than those illustrated herein.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

An implementation flowchart of a method for obtaining a centroid set representing a typical water body category provided in the embodiment of the present application may include:

Step S11: Acquiring water body sample data, the water body sample data includes remote sensing reflectance spectral data of several water surface sampling points;

The remote sensing reflectance spectrum data of the several water surface sampling points may be the remote sensing reflectance spectrum of each sampling point obtained through actual measurement. The remote sensing reflectance spectrum of optically complex water bodies is measured by the "above water surface measurement method". This measurement method is currently a general method for measuring the remote sensing reflectance spectrum of water bodies. Influence. And each site (that is, each sampling point) can obtain a remote sensing reflectance spectrum. The remote sensing reflectance spectrum refers to the situation that the reflectance of light changes with the wavelength. For example, the remote sensing reflectance spectrum can be 1nm (ie 1 nanometer) as an interval, including at least 551 remote sensing reflectances from 350nm to 900nm. Spectral, reflectance values are real numbers between 0 and 1. In the embodiment of the present invention, the spectral range and spectral resolution (that is, the interval between bands) of different sampling points are the same.

The remote sensing reflectance spectral data of the several water surface sampling points may also be images obtained by remote sensors covering optically complex water bodies. Usually, a remote sensing image with M rows, N columns, and L bands has a total of M*N=Num pixels, and the value of the L bands corresponding to each pixel can be regarded as the reflectance spectrum of the water surface point corresponding to the pixel .

Step S12: performing dimensionality reduction on the reflectance spectrum data of several bands of each water surface sampling point through principal component analysis transformation (i.e. PCA transformation), to obtain two principal component data of the reflectance spectrum of each water surface sampling point;

In the embodiment of the present invention, through PCA transformation, the reflectance spectral data of L bands at each sampling point are represented by two principal components. That is to say, in the embodiment of the present invention, before classification, the reflectance spectrum data of each sampling point is firstly reduced in dimension, which reduces the amount of data and the amount of calculation.

Step S13: Determine the number of categories C;

In the embodiment of the present invention, a fuzzy clustering algorithm may be used for fuzzy classification. Among them, fuzzy clustering is a kind of unsupervised classification, and the number of categories needs to be input in advance for unsupervised classification. Unsupervised classification of samples is only possible if the number of classes is known.

Among them, the number of categories C can be determined empirically by researchers.

Step S14: Execute the fuzzy classification process at least twice, and the distance measures based on each fuzzy classification process are different; the fuzzy classification process divides the several water surface sampling points according to the two principal component data of the reflectance spectrum of each water surface sampling point. For class C, get C initial centroid sets;

Each time the fuzzy classification process is performed, C initial centroid sets are obtained. Each initial centroid set represents a typical water body category, and each initial centroid set has several water body sample data similar or identical to the typical water body category.

Optionally, a fuzzy c-means clustering algorithm (fuzzy c-means algorithm, FCM) can be used to perform fuzzy classification on the water body sample data. Of course, in the embodiment of the present invention, it is not limited to using the FCM clustering algorithm for fuzzy classification, and other fuzzy clustering algorithms can also be used, such as the improved fuzzy clustering algorithm EFC-MD (Evolutionary fuzzy clustering with minkowski distances), as long as All fuzzy clustering algorithms based on distance measures are applicable to the embodiments of the present invention.

Distance is a simple and effective measure of data similarity. In the process of fuzzy classification, it is necessary to determine the weight coefficient of the pixel belonging to each category according to the distance between the pixel and the centroid pixel of each category.

In the embodiment of the present invention, the distance measure used in the j-th fuzzy classification process is different from the distance measure used in the previous j-1 fuzzy classification process, where j=1, 2, 3, ... J, J is The total number of times the fuzzy classification process is performed, J≥2. That is, the distance measures used in any two fuzzy classification processes are different.

Optionally, in the embodiment of the present invention, the distance measure can be used but not limited to the following: Euclidean distance (Euclidean distance, Euc), cosine distance (SAD), OPD divergence (orthogonalprojection divergence), TD divergence Degree (transformed divergence), Mahalanobis distance, etc.

Preferably, four fuzzy classification processes can be performed. Specifically, when the fuzzy classification process is performed four times, one distance measure is used for each time the fuzzy classification process is performed, and four distance measures are used for four times of fuzzy classification process execution. The selected distance measures can be respectively: Euclidean distance, cosine distance, OPD divergence and TD divergence. Certainly, in the embodiment of the present invention, it is not limited to these four types, and may be any four of the above five distance measures.

Step S15: Compute the intersection of the initial centroid sets corresponding to the i-th category obtained by each execution of the fuzzy classification process, and obtain the centroid set corresponding to the i-th category; where, i=1, 2, . . . , C.

Assuming that the C centroid sets obtained from the j-th fuzzy classification are U _j1 , U _j2 , ... U _jC , then the centroid set U _i corresponding to the i-th category is U _i = U _1i ∩U _2i ∩...∩ U _Ji , where, U _ji (j=1, 2, 3, ... J, J is the total number of fuzzy classification processes, J≥2) is the fuzzy classification for the jth time to obtain the i-th category corresponding Initial set of centroids.

By analyzing the environmental parameters of each sample in the centroid set corresponding to the i-th category, it can be determined what typical water body the centroid set corresponding to the i-th category represents. How to determine the typical water body represented by the centroid set corresponding to the i-th category belongs to common knowledge in the field, and will not be repeated here.

The reflectance spectrum in the centroid set corresponding to the i-th category is averaged to obtain the centroid spectrum representing a typical water body category.

The method for obtaining a centroid set representing a typical water body category provided by the embodiment of the present invention is to obtain water body sample data, perform dimensionality reduction on the reflectance spectrum data of each water surface sampling point to obtain two principal component data of the reflectance spectrum of each sampling point , determine the number of categories, carry out fuzzy classification to the several water surface sampling points according to the two principal component data of the reflectance spectrum of each water surface sampling point at least twice; wherein, the distance measure based on each fuzzy classification is different; Obtained by performing the fuzzy classification process for the second time, the initial centroid set corresponding to the i-th category is intersected to obtain the centroid set corresponding to the i-th category, where i=1, 2, ..., C. While improving the stability of the centroid set corresponding to the i-th category, it makes the centroid set more representative.

Preferably, since the determination of the number of categories C through experience is not objective enough, based on this, the embodiment of the present invention proposes that the number of categories C can be determined according to the sample data itself.

Optionally, an implementation flowchart for determining the number of categories C is shown in Figure 2, which may include:

Step S21: Calculate the BIC index of the several water surface sampling points under different classification numbers according to the two principal component data of the reflectance spectrum of each water surface sampling point;

BIC index is an index to evaluate the effectiveness of fuzzy clustering. In the embodiment of the present invention, the principal component data obtained after dimension reduction of each water surface sampling point is used to calculate the BIC index instead of the originally acquired reflectance spectrum data of the water surface sampling point, which reduces the calculation amount of the BIC index.

In the embodiment of the present invention, the number of categories changes one by one from 2 to 16, and the BIC index of the sample data is calculated every time the number of categories changes. When calculating the BIC index, the sample data after PCA transformation is used for calculation. The BIC index can be calculated according to the following formula:

$\mathrm{<span\; class="notranslate">\; BIC</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; no</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; log</span>}{\mathrm{<span\; class="notranslate">\; no</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; no</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; log</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; no</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; d</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; log</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; no</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; log</span>}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; no</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; m\; olg\; n</span>}$

where n _i is the number of samples in the i-th category. d is the data dimension of each water surface sampling point, which is equal to 2 here. m is the number of categories. n is the total number of water surface sampling points. ∑ _i is the maximum likelihood estimate of the variance of the ith category x _j is the j-th sample point in the i-th category, and C _i is the centroid of the i-th category.

Step S22: Determine the number of categories corresponding to the minimum BIC index as the number of categories C.

In the embodiment of the present invention, the optimal number of categories C is determined based on the samples themselves, which further enhances the stability of the obtained centroid set.

Optionally, another implementation flowchart for determining the number of categories C is shown in Figure 3, which may include:

Step S31: Calculate the Dunn'c index of the several water surface sampling points under different classification data according to the two principal component data of the reflectance spectrum of each water surface sampling point;

Dunn'c index is another index to evaluate the effectiveness of fuzzy clustering. In the embodiment of the present invention, the Dunn'c index is calculated by using the principal component data obtained after the dimension reduction of each water surface sampling point, instead of using the reflectance spectral data of the originally obtained water surface sampling point for calculation, which reduces the calculation of the Dunn'c index quantity.

In the embodiment of the present invention, the number of categories changes one by one from 2 to 16, and the Dunn'c index of the sample data is calculated every time the number of categories changes. When calculating the Dunn'c index, the sample data after PCA transformation are used for calculation.

Step S32: Determine the number of categories corresponding to the maximum Dunn'c index as the number of categories C.

In the embodiment of the present invention, the optimal number of categories C is determined based on the samples themselves, which further enhances the stability of the obtained centroid set.

Further, the acquired water body sample data also includes: the water body environmental variables of the several water surface sampling points, and the water body environmental variables include: water quality parameters and inherent optical quantities of different components of the water body; wherein, the water quality parameters may include: Chlorophyll a concentration, total particulate matter concentration, organic particulate matter concentration, inorganic particulate matter concentration, dissolved organic carbon concentration, etc. The intrinsic optical quantities of different components of the water body can include: beam attenuation coefficient of particles and CDOM, CDOM absorption coefficient, particle absorption coefficient, non-algae particle absorption coefficient, phytoplankton absorption coefficient, particle scattering coefficient, etc.

As shown in Figure 4, Figure 4 is another implementation flowchart of the method for obtaining a centroid set representing a typical water body category provided by the present application. On the basis of the embodiment shown in Figure 1, after obtaining the i-th category corresponding After the set of centroids, the method also includes:

Step S41: Display the water environment variable parameters of each sample in the centroid set corresponding to the i-th category;

Step S42: When receiving the elimination instruction triggered by the user, determine the target sample data according to the identification of the sample data carried in the elimination instruction, delete the target sample data, and re-execute the fuzzy classification process at least twice A step of.

It can be judged by the researchers (that is, users) based on experience whether the water environment variables of the sampling points in the i-th centroid set are obviously abnormal. When the researchers judge that the water environment variables of the sampling points in the i-th centroid set are obviously abnormal, they will trigger the deletion command to delete the sample data with obvious abnormalities, and the deletion command carries the identification mark of the sample data to be deleted.

In order to avoid the occurrence of foreign matter with the same spectrum, in the embodiment of the present invention, the researchers analyze and judge the measured environmental parameters corresponding to the samples in the same centroid set, and if the environmental parameters are abnormal, the samples of the sampling points with obvious abnormalities are eliminated data, re-execute step S14 and step S15 to obtain a new centroid set. There is no obvious abnormality in the water environmental variables up to the sampling points in each centroid set. The occurrence of different objects with the same spectrum is avoided, so that the type represented by the classified water body has physical meaning.

When a determination instruction triggered by the user is received, the centroid spectrum of the centroid set of the i-th category may be output. Other running results can also be output according to user requirements, such as the i-th centroid set, etc.

In the above embodiment, optionally, in order to make the determined centroid set have a wider application range, in the embodiment of the present invention, water body sample data of multiple water areas are collected. Specifically, in the embodiment of the present invention, the sampling area of the water body sample data may include but not limited to at least one of the following: inland waters, nearshore waters, and the like. Among them, inland waters may include: rivers, lakes, ponds, ponds, reservoirs and other waters. In the embodiment of the present invention, sampling may be performed only in one type of inland water area, or may be performed in more than two types of inland water areas.

For example, water body sample data can be collected in Taihu Lake, Three Gorges Reservoir, Dianchi Lake, Chaohu Lake and the mouth of the Yellow River.

Corresponding to the method embodiment, the present application also provides a device for obtaining a centroid set representing a typical water body category. A schematic structural diagram of the device for obtaining a centroid set representing a typical water body category provided by the present application is shown in FIG. 5 , which can be include:

Sample acquisition module 51, dimensionality reduction module 52, determination module 53, classification module 54 and centroid set acquisition module 55; wherein,

Sample acquiring module 51 is used for acquiring water body sample data, and described water body sample data comprises the remote sensing albedo spectrum data of several water surface sampling points;

The remote sensing reflectance spectrum data of the several water surface sampling points may be the remote sensing reflectance spectrum of each sampling point obtained through actual measurement. The remote sensing reflectance spectrum of optically complex water bodies is measured by the "above water surface measurement method". This measurement method is currently a general method for measuring the remote sensing reflectance spectrum of water bodies. Influence. And each site (that is, each sampling point) can obtain a remote sensing reflectance spectrum. The remote sensing reflectance spectrum refers to the situation that the reflectance of light changes with the wavelength. For example, the remote sensing reflectance spectrum can be 1nm (ie 1 nanometer) as an interval, including at least 551 remote sensing reflectances from 350nm to 900nm. Spectral, reflectance values are real numbers between 0 and 1. In the embodiment of the present invention, the spectral range and spectral resolution (that is, the interval between bands) of different sampling points are the same.

The remote sensing reflectance spectral data of the several water surface sampling points may also be images obtained by remote sensors covering optically complex water bodies. Usually, a remote sensing image with M rows, N columns, and L bands has a total of M*N=Num pixels, and the value of the L bands corresponding to each pixel can be regarded as the reflectance spectrum of the water surface point corresponding to the pixel .

Dimensionality reduction module 52 is used for carrying out dimensionality reduction to the reflectance spectrum data of several bands of each water surface sampling point through principal component analysis transformation, obtains two principal component data of the reflectance spectrum of each water surface sampling point;

In the embodiment of the present invention, through PCA transformation, the reflectance spectral data of L bands at each sampling point are represented by two principal components. That is to say, in the embodiment of the present invention, before classification, the reflectance spectrum data of each sampling point is firstly reduced in dimension, which reduces the amount of data and the amount of calculation.

Determining module 53 is used for determining category number C;

In the embodiment of the present invention, a fuzzy clustering algorithm may be used for fuzzy classification. Among them, fuzzy clustering is a kind of unsupervised classification, and the number of categories needs to be input in advance for unsupervised classification. Unsupervised classification of samples is only possible if the number of classes is known.

Among them, the number of categories C can be determined empirically by researchers.

The classification module 54 is used to perform the fuzzy classification process at least twice, and the distance measures based on each fuzzy classification process are different; the fuzzy classification process samples the several water surfaces according to the two principal component data of the reflectance spectrum of each water surface sampling point. The points are divided into C categories, and C initial centroid sets are obtained;

Each time the fuzzy classification process is performed, C initial centroid sets are obtained.

Optionally, a fuzzy c-means clustering algorithm (fuzzy c-means algorithm, FCM) can be used to perform fuzzy classification on the water body sample data. Of course, in the embodiment of the present invention, it is not limited to using the FCM clustering algorithm for fuzzy classification, and other fuzzy clustering algorithms can also be used, such as the improved fuzzy clustering algorithm EFC-MD (Evolutionary fuzzy clustering with minkowski distances), as long as All fuzzy clustering algorithms based on distance measures are applicable to the embodiments of the present invention.

Distance is a simple and effective measure of data similarity. In the process of fuzzy classification, it is necessary to determine the weight coefficient of the pixel belonging to each category according to the distance between the pixel and the centroid pixel of each category.

In the embodiment of the present invention, the distance measure used in the j-th fuzzy classification process is different from the distance measure used in the previous j-1 fuzzy classification process, where j=1, 2, 3, ... J, J is The total number of times the fuzzy classification process is performed, J≥2. That is, the distance measures used in any two fuzzy classification processes are different.

Optionally, in the embodiment of the present invention, the distance measure can be used but not limited to the following: Euclidean distance (Euclidean distance, Euc), cosine distance (SAD), OPD divergence (orthogonalprojection divergence), TD divergence Degree (transformed divergence), Mahalanobis distance, etc.

Preferably, four fuzzy classification processes can be performed. Specifically, when the fuzzy classification process is performed four times, one distance measure is used for each time the fuzzy classification process is performed, and four distance measures are used for four times of fuzzy classification process execution. The selected distance measures can be respectively: Euclidean distance, cosine distance, OPD divergence and TD divergence. Certainly, in the embodiment of the present invention, it is not limited to these four types, and may be any four of the above five distance measures.

The centroid set acquisition module 55 is used to obtain the intersection of the initial centroid sets corresponding to the i-th category obtained by each execution of the fuzzy classification process, and obtain the centroid set corresponding to the i-th category; wherein, i=1, 2, ... ..., C.

Assuming that the C centroid sets obtained from the j-th fuzzy classification are U _j1 , U _j2 , ... U _jC , then the centroid set U _i corresponding to the i-th category is U _i = U _1i ∩U _2i ∩...∩ U _Ji , where, U _ji (j=1, 2, 3, ... J, J is the total number of fuzzy classification processes, J≥2) is the fuzzy classification for the jth time to obtain the i-th category corresponding Initial set of centroids.

By analyzing the environmental parameters of each sample in the centroid set corresponding to the i-th category, it can be determined what typical water body the centroid set corresponding to the i-th category represents. How to determine the typical water body represented by the centroid set corresponding to the i-th category belongs to common knowledge in the field, and will not be repeated here.

The reflectance spectrum in the centroid set corresponding to the i-th category is averaged to obtain the centroid spectrum representing a typical water body category.

The device for obtaining a centroid set representing a typical water body type provided by an embodiment of the present invention obtains water body sample data, and performs dimensionality reduction on the reflectance spectrum data of each water surface sampling point to obtain two principal component data of the reflectance spectrum of each sampling point , determine the number of categories, carry out fuzzy classification to the several water surface sampling points according to the two principal component data of the reflectance spectrum of each water surface sampling point at least twice; wherein, the distance measure based on each fuzzy classification is different; Obtained by performing the fuzzy classification process for the second time, the initial centroid set corresponding to the i-th category is intersected to obtain the centroid set corresponding to the i-th category, where i=1, 2, ..., C. While improving the stability of the centroid set corresponding to the i-th category, it makes the centroid set more representative.

Optionally, a schematic structural diagram of the determining module 53 is shown in FIG. 6, which may include:

The first calculating unit 61 and the first determining unit 62; wherein,

The first calculation unit 61 is used to calculate the BIC index of the several water surface sampling points under different classification numbers according to the two principal component data of the reflectance spectrum of each water surface sampling point;

BIC index is an index to evaluate the effectiveness of fuzzy clustering. In the embodiment of the present invention, the principal component data obtained after dimension reduction of each water surface sampling point is used to calculate the BIC index instead of the originally acquired reflectance spectrum data of the water surface sampling point, which reduces the calculation amount of the BIC index.

In the embodiment of the present invention, the number of categories changes one by one from 2 to 16, and the BIC index of the sample data is calculated every time the number of categories changes. When calculating the BIC index, the sample data after PCA transformation is used for calculation.

The first determining unit 62 is configured to determine the category number C corresponding to the minimum BIC index as the category number C.

In the embodiment of the present invention, the optimal number of categories C is determined based on the samples themselves, which further enhances the stability of the obtained centroid set.

Optionally, another structural schematic diagram of the determination module 53 is shown in FIG. 7, which may include:

The second calculating unit 71 and the second determining unit 72; wherein,

The second calculation unit 71 is used to calculate the Dunn'c index of the several water surface sampling points under different classification data according to the two principal component data of the reflectance spectrum of each water surface sampling point;

Dunn'c index is another index to evaluate the effectiveness of fuzzy clustering. In the embodiment of the present invention, the Dunn'c index is calculated by using the principal component data obtained after the dimension reduction of each water surface sampling point, instead of using the reflectance spectral data of the originally obtained water surface sampling point for calculation, which reduces the calculation of the Dunn'c index quantity.

In the embodiment of the present invention, the number of categories changes one by one from 2 to 16, and the Dunn'c index of the sample data is calculated every time the number of categories changes. When calculating the Dunn'c index, the sample data after PCA transformation are used for calculation.

The second determination unit 72 is used to determine the number of categories corresponding to the maximum Dunn'c index as the number of categories C.

In the embodiment of the present invention, the optimal number of categories C is determined based on the samples themselves, which further enhances the stability of the obtained centroid set.

Further, in the embodiment of the present invention, the water body sample data may also include: the water body environmental variables of the several water surface sampling points, and the water body environmental variables include: water quality parameters and inherent optical quantities of different components of the water body; wherein, the water quality The parameters may include: chlorophyll a concentration, total particulate matter concentration, organic particulate matter concentration, inorganic particulate matter concentration, dissolved organic carbon concentration, and the like. The intrinsic optical quantities of different components of the water body can include: beam attenuation coefficient of particles and CDOM, CDOM absorption coefficient, particle absorption coefficient, non-algae particle absorption coefficient, phytoplankton absorption coefficient, particle scattering coefficient, etc.

On the basis of the embodiment shown in Figure 5, another structural schematic diagram of the device for obtaining a centroid set representing a typical water body category provided by the application of the present invention is shown in Figure 8, and may also include:

Display module 81 and delete module 82; Wherein,

The display module 81 is used to display the water body environment variable parameters of each sample in the centroid set corresponding to the i category;

The deletion module 82 is used to determine the target sample data according to the identification of the sample data carried in the removal instruction when receiving the removal instruction triggered by the user, delete the target sample data, and generate a trigger instruction to instruct the classification module The step of performing the fuzzy classification process at least twice is re-performed.

It can be judged by the researchers (that is, users) based on experience whether the water environment variables of the sampling points in the i-th centroid set are obviously abnormal. When the researchers judge that the water environment variables of the sampling points in the i-th centroid set are obviously abnormal, they will trigger the deletion command to delete the sample data with obvious abnormalities, and the deletion command carries the identification mark of the sample data to be deleted.

In order to avoid the occurrence of different objects with the same spectrum, in the embodiment of the present invention, the researchers analyze and judge the measured environmental parameters corresponding to the samples in the same centroid set. If the environmental parameters are abnormal, the classification module 54 and the acquisition of the centroid set will be retriggered. Module 55 works to get a new set of centroids. There is no obvious abnormality in the water environmental variables up to the sampling points in each centroid set. The occurrence of different objects with the same spectrum is avoided, so that the type represented by the classified water body has physical meaning.

In the above embodiment, optionally, the sample acquisition module is specifically configured to acquire water body sample data, and the water body sample data includes remote sensing reflectance spectral data of several water surface sampling points; the sampling area of the water body sample data can be Including but not limited to at least one of the following: inland waters and coastal waters, etc. Among them, inland waters may include: rivers, lakes, ponds, ponds, reservoirs and other waters. In the embodiment of the present invention, sampling may be performed only in one type of inland water area, or may be performed in more than two types of inland water area.

For example, water body sample data can be collected in Taihu Lake, Three Gorges Reservoir, Dianchi Lake, Chaohu Lake and the mouth of the Yellow River.

Those skilled in the art can clearly understand that for the convenience and brevity of the description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

The above description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention will not be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (10)

1. A method of obtaining a set of centroids representative of a typical water body class, comprising:

acquiring water body sample data, wherein the water body sample data comprises remote sensing reflectivity spectrum data of a plurality of water surface sampling points;

performing dimensionality reduction on the reflectivity spectrum data of a plurality of wave bands of each water surface sampling point through principal component analysis and transformation to obtain two principal component data of the reflectivity spectrum of each water surface sampling point;

determining the number C of categories;

performing fuzzy classification processes at least twice, wherein distance measures based on the fuzzy classification processes are different in each time; the fuzzy classification process divides the water surface sampling points into C types according to two principal component data of the reflectivity spectrum of each water surface sampling point to obtain C initial mass center sets;

solving an intersection of the initial mass center set obtained by executing the fuzzy classification process and corresponding to the ith category to obtain a mass center set corresponding to the ith category; wherein i is 1, 2, … …, C.

2. The method of claim 1, wherein the determining the number of classes C comprises:

calculating the BIC indexes of the water surface sampling points under different classification numbers according to two main component data of the reflectivity spectrum of each water surface sampling point;

and determining the classification number corresponding to the minimum BIC index as the class number C.

3. The method of claim 1, wherein the determining the number of classes C comprises:

calculating Dunn' c indexes of the plurality of water surface sampling points under different classification data according to two main component data of the reflectivity spectrum of each water surface sampling point;

and determining the classification number corresponding to the maximum Dunn' C index as the classification number C.

4. The method of claim 1, wherein the water sample data further comprises: the water environment variable of a plurality of surface sampling points comprises: water quality parameters and inherent optical quantities of different components of the water body; after obtaining the set of centroids corresponding to the ith class, the method further comprises:

displaying the water body environment variable parameters of all samples in the centroid set corresponding to the ith category;

when a removing instruction triggered by a user is received, determining target sample data according to the identification mark of the sample data carried in the removing instruction, deleting the target sample data, and executing the step of executing the fuzzy classification process at least twice again.

5. The method of claim 1, wherein the sampling region of water body sample data comprises at least one of: inland waters and near-shore waters.

6. An apparatus for acquiring a set of centroids representative of a typical class of water, comprising:

the system comprises a sample acquisition module, a data acquisition module and a data acquisition module, wherein the sample acquisition module is used for acquiring water body sample data which comprises remote sensing reflectivity spectrum data of a plurality of water surface sampling points;

the dimension reduction module is used for reducing the dimension of the reflectivity spectrum data of a plurality of wave bands of each water surface sampling point through principal component analysis and transformation to obtain two principal component data of the reflectivity spectrum of each water surface sampling point;

a determining module for determining the number of categories C;

the classification module is used for executing fuzzy classification processes at least twice, and the fuzzy classification processes in each time are different in distance measure; the fuzzy classification process classifies the acquired water surface sampling points into C types according to two principal component data of the reflectivity spectrum of each water surface sampling point to obtain C initial mass center sets;

the centroid set acquisition module is used for solving an intersection of the initial centroid set obtained by executing the fuzzy classification process and corresponding to the ith category to obtain a centroid set corresponding to the ith category; wherein i is 1, 2, … …, C.

7. The apparatus of claim 6, wherein the determining module comprises:

the first calculation unit is used for calculating the BIC indexes of the water surface sampling points under different classification numbers according to two main component data of the reflectivity spectrum of each water surface sampling point;

and the first determining unit is used for determining the classification number corresponding to the minimum BIC index as the category number C.

8. The apparatus of claim 6, wherein the determining module comprises:

the second calculation unit is used for calculating Dunn' c indexes of the plurality of water surface sampling points under different classification data according to two main component data of the reflectivity spectrum of each water surface sampling point;

and the second determining unit is used for determining the classification number corresponding to the maximum Dunn' C index as the classification number C.

9. The apparatus of claim 6, wherein the water body sample data further comprises: the water environment variable of a plurality of surface sampling points comprises: water quality parameters and inherent optical quantities of different components of the water body; the device further comprises:

the display module is used for displaying the water body environment variable parameters of all samples in the centroid set corresponding to the ith category;

and the deleting module is used for determining target sample data according to the identification identifier of the sample data carried in the removing instruction when a removing instruction triggered by a user is received, deleting the target sample data, and generating a triggering instruction to instruct the classifying module to execute the step of executing the fuzzy classifying process at least twice again.

10. The device of claim 6, wherein the sample acquisition module is specifically configured to acquire water sample data comprising remote sensing reflectance spectral data of a plurality of water surface sampling points; the sampling region of the water body sample data comprises at least one of the following regions: inland waters and near-shore waters.