CN103926203B - A kind of for the probabilistic Spectral Angle Mapping method of object spectrum - Google Patents

Abstract

The present invention proposes a kind of for the probabilistic Spectral Angle Mapping method of object spectrum, comprising: obtain test spectral and reference spectra; Utilize described test spectral and reference spectra to calculate SPECTRAL DIVERSITY amount, and build according to described SPECTRAL DIVERSITY amount and with the vector of described test spectral, there is same dimension and the equal SPECTRAL DIVERSITY of each component size is vectorial; Utilize the spectral modeling of described SPECTRAL DIVERSITY vector calculation in the uncertain situation of object spectrum between described test spectral and reference spectra; Spectral Angle Mapping is carried out according to described spectral modeling.Adopt the method that the present invention proposes, in the uncertain situation of object spectrum, the spectral modeling between test spectral and reference spectra is obtained by SPECTRAL DIVERSITY amount, Spectral Angle Mapping is carried out according to the spectral modeling obtained, overcome the precision that the impact brought because object spectrum is uncertain improves Objects recognition, to object spectrum uncertainty, there is better applicability.

Description

Spectral angle mapping method for ground object spectral uncertainty

Technical Field

The invention relates to the technical field of remote sensing, in particular to a spectral angle mapping method aiming at surface feature spectral uncertainty, which is used for hyperspectral remote sensing mineral mapping and target identification.

Background

The hyperspectral remote sensing data provides a large amount of spectral information of the ground objects, and fine classification and quantitative remote sensing of the ground objects are facilitated. The spectral angle mapping algorithm (SAM) is an algorithm based on the overall similarity of spectral curves, and is widely applied to hyperspectral remote sensing information classification.

The similarity of the whole spectral curve is calculated by the spectral angle mapping algorithm, the similarity is a global description index, and the ground features with similar spectral curves are poorly distinguished. At present, the spectral angle mapping algorithm is improved mainly by setting the local interval weight, selecting the wave band combination, introducing the kernel function and other methods. Specific methods of improvement can be found in reference 1: hyperspectral mineral mapping method based on weight spectral angle mapping [ J ] Spectroscopy and spectral analysis, 2011,31(8): 2200-: luyinliang, li shaohu, wangke, etc. wheat yield monitoring studies based on improved spectral angle algorithms [ J ]. Sinkiang agricultural sciences, 2011,48(001):1-5 and literature 3: GuY, WangC, WangS, et al, Kernel-based-estimated-for-target-determined-hyperspectral [ J ]. Pattern recognitions letters,2011,32(2): 114-. However, the uncertainty of the surface feature spectrum often causes a certain degree of difference between the same surface feature spectrum, which affects the identification precision of the surface feature, and also has a certain effect on the surface feature identification effect of the spectral angle mapping algorithm.

Disclosure of Invention

Technical problem to be solved

The invention aims to provide a spectral angle mapping method aiming at the spectral uncertainty of a ground object, and the method is used for solving the problem that the spectral uncertainty of the ground object is not considered in the conventional spectral angle mapping method.

(II) technical scheme

In order to achieve the above object, the present invention provides a method for mapping spectrum angles of uncertainty of a surface feature spectrum, comprising:

acquiring a test spectrum and a reference spectrum;

calculating a spectrum difference by using the test spectrum and the reference spectrum, and constructing a spectrum difference vector which has the same dimension as the vector of the test spectrum and has the same component size as the vector of the test spectrum according to the spectrum difference;

calculating a spectral angle between the test spectrum and a reference spectrum with the feature spectral uncertainty using the spectral difference vector;

and carrying out spectral angle mapping according to the spectral angle.

Preferably, the acquiring of the test spectrum and the reference spectrum specifically comprises:

acquiring the test spectrum by utilizing a hyperspectral image, actually measured spectrum data or a spectrum library;

and acquiring the reference spectrum by utilizing a hyperspectral image, actually measured spectrum data or a spectrum library.

Preferably, the calculating the spectral difference using the test spectrum and the reference spectrum specifically includes:

acquiring spectrum vectors of the test spectrum and the reference spectrum;

calculating a spectrum difference n according to the spectrum vectors of the obtained test spectrum and the reference spectrum, wherein the formula is as follows:

$n=\frac{(\stackrel{\→}{t}\·\stackrel{\→}{r})\×\mathrm{\Σ}{t}_i-\mathrm{\Σ}{r}_i\×\mathrm{\Σ}{t}_i^2}{\mathrm{\Σ}{t}_i\×\mathrm{\Σ}{r}_i-\mathrm{nb}\×(\stackrel{\→}{t}\·\stackrel{\→}{r})}$

wherein, $\stackrel{\→}{t}=(t_1,t_2,...,t_{\mathrm{nb}})$ a vector representing the spectrum of the test spectrum, $\stackrel{\→}{r}=(r_1,r_2,...,r_{\mathrm{nb}})$ representing the reference spectral vector, nb the dimensionality of the spectral vector, and i any integer between 1 and nb.

Preferably, the calculating the spectrum angle between the test spectrum and the reference spectrum under the uncertainty of the feature spectrum by using the spectrum difference vector specifically includes:

calculating a cosine value of a spectrum angle between the test spectrum and the reference spectrum under the condition of the uncertainty of the ground feature spectrum according to the spectrum difference vector, wherein the formula is as follows:

$\cos \mathrm{\α}\text{'}=\frac{(\stackrel{\→}{t}+\stackrel{\→}{c})\·\stackrel{\→}{r}}{\left|\right|\stackrel{\→}{t}+\stackrel{\→}{c}\left|\right|\×\left|\right|\stackrel{\→}{r}\left|\right|}$

wherein, $\stackrel{\→}{t}=(t_1,t_2,...,t_{\mathrm{nb}})$ a vector representing the spectrum of the test spectrum, $\stackrel{\→}{r}=(r_1,r_2,...,r_{\mathrm{nb}})$ a reference spectral vector is represented that is,representing a spectral difference vector, n being a spectral difference, α' being a spectral angle between the test spectral vector and the reference spectral vector in case of a spectral difference;

and calculating the spectrum angle between the test spectrum and the reference spectrum under the condition of uncertainty of the ground feature spectrum according to the cosine value of the spectrum angle based on the maximum principle of the cosine of the spectrum angle between the spectra of the same ground feature.

Preferably, the test spectrum is obtained by removing four water vapor strong absorption wave bands in a solar reflection spectrum range of 350-2500 nm.

Preferably, the reference spectrum is obtained by removing the four water vapor strong absorption wave bands in a solar reflection spectrum range of 350-2500 nm.

Preferably, the four water vapor strong absorption bands include: 900 to 990nm, 1100 to 1190nm, 1300 to 1520nm and 1750 to 2080 nm.

(III) advantageous effects

According to the spectral angle mapping method for the ground feature spectrum uncertainty, the spectrum difference quantity can be used for effectively representing the spectrum difference of the same ground feature, the influence of the ground feature spectrum uncertainty is overcome, the ground feature identification precision is improved, and the method has better applicability to the ground feature spectrum uncertainty.

Drawings

FIG. 1 is a flow chart of a method for mapping spectral angles of spectral uncertainty of a surface feature according to the present invention;

FIG. 2 is a reference spectrum of kaolinite selected from the USGS standard spectra library in an example of the invention;

FIG. 3 is a test spectrum of kaolinite selected from the USGS standard spectra library in an example of the invention.

Detailed Description

The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.

Fig. 1 is a flowchart of a spectral angle mapping method for the uncertainty of a surface feature spectrum according to the present invention, and as shown in the figure, the embodiment of the present invention includes the following steps:

s101, acquiring a test spectrum and a reference spectrum; the method mainly comprises the step of obtaining the test spectrum and the reference spectrum by utilizing a hyperspectral image, actually measured spectrum data or a spectrum library.

In the embodiment of the invention, the reference spectrum of the kaolinite is selected from the USGS standard spectrum library, and as shown in figure 2, four wave bands with strong water vapor absorption are removed: 900-990 nm, 1100-1190 nm, 1300-1520 nm, 1750-2080 nm; the kaolinite test spectrum was selected from the USGS standard spectral library, as shown in fig. 3, with the four bands of strong water vapor absorption removed: 900 to 990nm, 1100 to 1190nm, 1300 to 1520nm, 1750 to 2080 nm.

S102, calculating a spectrum difference by using the test spectrum and the reference spectrum, and constructing a spectrum difference vector which has the same dimension as the vector of the test spectrum and has the same component size according to the spectrum difference, specifically comprising:

in the embodiment of the invention, the spectrum difference n is calculated by acquiring the spectrum vectors of the test spectrum and the reference spectrum and according to the acquired spectrum vectors of the test spectrum and the reference spectrum, and the formula is as follows:

$n=\frac{(\stackrel{\→}{t}\·\stackrel{\→}{r})\×\mathrm{\Σ}{t}_i-\mathrm{\Σ}{r}_i\×\mathrm{\Σ}{t}_i^2}{\mathrm{\Σ}{t}_i\×\mathrm{\Σ}{r}_i-\mathrm{nb}\×(\stackrel{\→}{t}\·\stackrel{\→}{r})}$

wherein, $\stackrel{\→}{t}=(t_1,t_2,...,t_{\mathrm{nb}})$ a vector representing the spectrum of the test spectrum, $\stackrel{\→}{r}=(r_1,r_2,...,r_{\mathrm{nb}})$ representing the reference spectral vector, nb the dimensionality of the spectral vector, and i any integer between 1 and nb. The calculated spectral difference n value was 0.0899.

S103, calculating a spectrum angle between the test spectrum and the reference spectrum under the condition of the uncertainty of the surface feature spectrum by using the spectrum difference vector, and specifically comprising the following steps:

in the embodiment of the present invention, the spectral difference n (n is 0.0899) obtained in step S103 is obtained, and the cosine of the spectral angle between the test spectrum and the reference spectrum under the circumstance of uncertainty of the feature spectrum is calculated according to the spectral difference vector, where the formula is as follows:

$\cos \mathrm{\α}\text{'}=\frac{(\stackrel{\→}{t}+\stackrel{\→}{c})\·\stackrel{\→}{r}}{\left|\right|\stackrel{\→}{t}+\stackrel{\→}{c}\left|\right|\×\left|\right|\stackrel{\→}{r}\left|\right|}$

wherein, $\stackrel{\→}{t}=(t_1,t_2,...,t_{\mathrm{nb}})$ a vector representing the spectrum of the test spectrum, $\stackrel{\→}{r}=(r_1,r_2,...,r_{\mathrm{nb}})$ a reference spectral vector is represented that is,representing a spectral difference vector, n being a spectral difference, α' being a spectral angle between the test spectral vector and the reference spectral vector in case of a spectral difference;

and calculating the spectrum angle between the test spectrum and the reference spectrum under the condition of uncertainty of the ground feature spectrum according to the cosine value of the spectrum angle based on the maximum principle of the cosine of the spectrum angle between the spectra of the same ground feature.

The spectral angle α' between the test spectrum and the reference spectrum under the uncertainty of the surface feature spectrum, which was obtained in the example of the present invention, was 0.0069 radian. The spectral angle calculated without considering the amount of spectral difference was 0.0177 radians, and it can be seen that the improved spectral angle calculation results in a greater improvement.

And S104, carrying out spectrum angle mapping according to the spectrum angle, and carrying out spectrum angle mapping according to the improved spectrum angle, thereby effectively improving the accuracy of ground feature identification.

By adopting the spectral angle mapping method aiming at the spectral uncertainty of the ground feature, the spectral angle between the test spectrum and the reference spectrum is obtained under the condition of the spectral uncertainty of the ground feature through the spectral difference, and the spectral angle mapping is carried out according to the obtained spectral angle, so that the influence caused by the spectral uncertainty of the ground feature is overcome, the accuracy of ground feature identification is improved, and the method has better applicability to the spectral uncertainty of the ground feature.

The above embodiments are only for illustrating the invention and are not to be construed as limiting the invention, and those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention, therefore, all equivalent technical solutions also belong to the scope of the invention, and the scope of the invention is defined by the claims.

Claims (6)

1. A spectral angle mapping method for the uncertainty of a ground feature spectrum is characterized by comprising the following steps:

acquiring a test spectrum and a reference spectrum;

calculating a spectrum difference by using the test spectrum and the reference spectrum, and constructing a spectrum difference vector which has the same dimension as the vector of the test spectrum and has the same component size as the vector of the test spectrum according to the spectrum difference;

calculating a spectral angle between the test spectrum and a reference spectrum with the feature spectral uncertainty using the spectral difference vector;

carrying out spectrum angle mapping according to the spectrum angle;

the calculating the spectral difference by using the test spectrum and the reference spectrum specifically comprises:

acquiring spectrum vectors of the test spectrum and the reference spectrum;

calculating a spectrum difference n according to the spectrum vectors of the obtained test spectrum and the reference spectrum, wherein the formula is as follows:

$n=\frac{(\stackrel{\→}{t}\·\stackrel{\→}{r})\×{\mathrm{\Σt}}_i-{\mathrm{\Σr}}_i\×{\mathrm{\Σt}}_i^2}{{\mathrm{\Σt}}_i\×{\mathrm{\Σr}}_i-nb\×(\stackrel{\→}{t}\·\stackrel{\→}{r})}$

wherein, $\stackrel{\→}{t}=(t_1,t_2,...,t_{nb})$ a vector representing the spectrum of the test spectrum, $\stackrel{\→}{r}=(r_1,r_2,...,r_{nb})$ representing the reference spectral vector, nb the dimensionality of the spectral vector, and i any integer between 1 and nb.

2. The method of claim 1, wherein the acquiring the test spectrum and the reference spectrum specifically comprises:

acquiring the test spectrum by utilizing a hyperspectral image, actually measured spectrum data or a spectrum library;

and acquiring the reference spectrum by utilizing a hyperspectral image, actually measured spectrum data or a spectrum library.

3. The method of claim 1, wherein the calculating the spectral angle between the test spectrum and the reference spectrum with the surface feature spectral uncertainty using the spectral difference vector comprises:

calculating a cosine value of a spectrum angle between the test spectrum and the reference spectrum under the condition of the uncertainty of the ground feature spectrum according to the spectrum difference vector, wherein the formula is as follows:

${\mathrm{cos\α}}^{,}=\frac{(\stackrel{\→}{t}+\stackrel{\→}{c})\·\stackrel{\→}{r}}{\left|\right|\stackrel{\→}{t}+\stackrel{\→}{c}\left|\right|\×\left|\right|\stackrel{\→}{r}\left|\right|}$

wherein, $\stackrel{\→}{t}=(t_1,t_2,...,t_{nb})$ a vector representing the spectrum of the test spectrum, $\stackrel{\→}{r}=(r_1,r_2,...,r_{nb})$ a reference spectral vector is represented that is,representing a spectral difference vector, n being a spectral difference, α' being a spectral angle between the test spectral vector and the reference spectral vector in case of a spectral difference;

and calculating the spectrum angle between the test spectrum and the reference spectrum under the condition of uncertainty of the ground feature spectrum according to the cosine value of the spectrum angle based on the maximum principle of the cosine of the spectrum angle between the spectra of the same ground feature.

4. The method according to claim 1, wherein the test spectrum is a spectrum obtained by removing four strong water vapor absorption bands within a solar reflection spectrum range of 350-2500 nm.

5. The method according to claim 1 or 4, wherein the reference spectrum is a spectrum obtained by removing the four strong water vapor absorption bands within a solar reflection spectrum range of 350-2500 nm.

6. The method of claim 5, wherein the four strong water vapor absorption bands comprise: 900 to 990nm, 1100 to 1190nm, 1300 to 1520nm and 1750 to 2080 nm.