CN1595107A - Multiple optical spectrum alteration zoning method based on wavelet packet variation - Google Patents

Abstract

This invention is multiple spectrum etching zonation method based on small wave packet vary and the small wave packet used can not only recursive analyze low-pass filter, but also can do high-pass filter and in this way can effectively extract etching information from the detail subcomponent of small wave packet. So this method can get etching band or etching rock distribution information and extract detail structure information of rock etching and further analyze the etching band. It separately makes small wave analysis to two wave band images with proper layer number and aims to choose different joint coefficient and especially detail variable coefficient and calculate the differential variables of chooses joint of two wave bands. It uses joint differential variable sum and recreate etching image and closely classify the etching band of the scatter diagram by use of recreation image and any original image.

Description

Multispectral alteration based on the wavelet packet variation divides the band method

Technical field

The present invention relates to a kind of multispectral alteration and divide the band method, especially relate to a kind of multispectral alteration and divide the band method based on the wavelet packet variation.

Background technology

Wall rock alteration is the important geology phenomenon, also is important indicator for deposit.Adopting with TM is in the multispectral remote sensing image of representative, because multispectral image wave band bandwidth is bigger, the wave spectrum difference of altered rock and background rock is significantly reduction owing to the average effect between the image band wide spectral range, add the influence of surface rock lacquer, lichens, soil, plant, alteration Information generally only is present in the background of image as a kind of weak signal on remote sensing images.How to extract or be partitioned into from background alteration Information from remote sensing image is a difficult problem of constantly exploring in resources remote sensing image analysis and the application always.Flame Image Process that ratioing technigue and principal component analysis (PCA) etc. are commonly used and information extracting method are when the more weak and interference noise of information is strong, and effect is often undesirable.On the other hand, the alteration of dissimilar and varying strength is at TM, ETM ⁺Generally all show as a kind of comprehensive effect on the wide band image, information commonly used strengthens and extracting method is generally all done it as a whole on image they and background difference and come, and the detailed information that can't spy on relevant alteration zone inner structure and composition.

Wavelet transformation and multiresolution analysis are widely used in Flame Image Process with in analyzing, and wavelet analysis is mainly used in the image smoothing of image and noise removing, multiple dimensioned edge enhancing and detection, image recovery, texture analysis, image segmentation and image recognition, Image Data Compression, data fusion etc. in the remote sensing image processing.General wavelet transformation only carries out the recurrence decomposition to the impulse response of low-pass filter.

What remote sensing images obtained and write down is the energy of atural object radiation between different wave length, and what image texture reflected is the spatial variations of brightness or gray scale.The lithological change or the Extract Mineralized Alteration that become the ore deposit Diagn to follow and cause are undergone mutation its background object spectrum with respect to not alteration.In corresponding spectrum range, non-alterated rocks category information has bigger similarity in two band images because of characteristic spectrum difference is less, when utilizing wavelet decomposition, its information focuses mostly on the summary component, and the tool alteration Information is reflected on the details component because of spectral signature has than big-difference more.Therefore general wavelet transformation is difficult to carry out the extraction of alteration Information and cut apart.

Summary of the invention

The present invention is difficult to carry out the extraction of alteration Information and the shortcoming of cutting apart in order to overcome wavelet transformation, provides a kind of and utilizes the multispectral alteration of wavelet packet variation to divide the band method.

The WAVELET PACKET DECOMPOSITION that the present invention utilizes is with respect to general wavelet transformation, and it not only does the recurrence decomposition to low-pass filter, also Hi-pass filter is made recurrence simultaneously and decomposes, and can extract alteration Information effectively from the details component of WAVELET PACKET DECOMPOSITION like this.Therefore, the multispectral alteration that the present invention has designed based on the wavelet packet variation divides the band method to obtain alteration zone or alterated rocks distributed intelligence, extracts the fine structure information of rock ore deposit alteration, and alteration zone is done further segmentation.

Multispectral alteration based on the wavelet packet variation of the present invention divides the band method, comprises the steps:

1. according to the multispectral image that obtains, select suitable wavelet packet basis;

2. with selected wavelet packet basis, respectively the image of two wave bands is carried out the WAVELET PACKET DECOMPOSITION of the suitable number of plies;

3. different node coefficients, especially details component coefficient on purpose selecting are 2. found the solution the difference component of two wave band selected node coefficients; Realization is subdued or is suppressed the coefficient of dissociation of non-alteration Information;

4. the difference component sum of node is rebuild the alteration image in utilizing 3.;

5. utilizing 4., the scatter diagram of reconstructed image and arbitrary original image carries out the alteration zone segmentation.

Multispectral alteration based on the wavelet packet variation of the present invention divides the band method, and used wave band is at different alteration types, and to hydroxyl class and carbonate mineral alteration, described two wave band images are TM/ETM ⁺Five wave bands (1550～1750nm) and seven wave bands (2080/2090～2350nm).

Multispectral alteration based on the wavelet packet variation of the present invention divides the band method, wherein alteration component (AC:Altered Component) can be expressed as the difference component of different-waveband WAVELET PACKET DECOMPOSITION coefficient (X), and the difference component of the WAVELET PACKET DECOMPOSITION coefficient of different-waveband is calculated by following formula:

$\mathrm{AC}=\underset{i}{\mathrm{\Σ}}[\underset{j=0}{\overset{2^{2\×i}-1}{\mathrm{\Σ}}}{\left(X_{i,j}\right)}_{\mathrm{ima}{\mathrm{ge}}_1}-\underset{j=0}{\overset{2^{2\×i}}{\mathrm{\Σ}}}{\left(X_{i,j}\right)}_{{\mathrm{image}}_2}]---i=\mathrm{1,2},\·\·\·,n$

Advantage of the present invention, the present invention has utilized wavelet package transforms not only low-pass filter to be decomposed, and also Hi-pass filter is made recurrence simultaneously and decomposes, and can extract alteration Information effectively from the details component of WAVELET PACKET DECOMPOSITION like this.Therefore, method of the present invention can be obtained alteration zone or alterated rocks distributed intelligence, extracts the fine structure information of rock ore deposit alteration, and alteration zone is done further segmentation.

Multispectral alteration based on the wavelet packet variation of the present invention divides other details of band method and the characteristics can be cheer and bright by reading the embodiment that hereinafter encyclopaedizes in conjunction with the accompanying drawings.

Description of drawings

Fig. 1 is beach, Bake test site TM7,4,1 color composite image;

Fig. 2 is based on the alteration Information figure of wavelet packet variation;

The two-dimentional scatter diagram of Fig. 3 TM5 and alteration Information figure;

Fig. 4 is the alteration zone segmentation;

Fig. 5 is the details of alteration distribution in the red frame among Fig. 1.

Embodiment

The altered mineral spectral signature embodies a concentrated expression of the characteristic absorption that altered mineral appears in 1500～2500nm zone, promptly is equivalent to TM/ETM ⁺((2080/2090～2350nm) spectrum and image value are with respect to the not variation of the background of alteration for 1550～1750nm), seven wave bands for five wave bands.Because the spectrum that alteration produced of dissimilar and varying strength and the variation of brightness of image have its different feature on TM5 and TM7 wave band, so can utilize wavelet packet to image detail and small identification.

Utilize wavelet package transforms with ETM ⁺The 5th, 7 band image signals resolve into summary (low frequency) component and the details component of a series of different scales.In corresponding spectrum range, non-alterated rocks category information has bigger similarity in two band images because of characteristic spectrum difference is less, its information focuses mostly on the summary component, and the tool alteration Information is reflected on the details component because of spectral signature has than big-difference more.Adopt wavelet package transforms under different resolution, repeatedly image to be decomposed, form the sub-band images of different directions, different resolution, realization is to the focusing in image information different levels and orientation, strengthen the similarity and the difference of information under the different scale, thereby realize the information of image small detail or faint variation is detected.The alteration Information frequency distribution has certain dynamic range, shows as high-frequency signal in low frequency component, is low frequency signal in the details component.In the details component, include high frequency information such as road, water system toward contact, their curve of spectrum is often straight and do not have changing features, wavelet packet component (coefficient of dissociation) to two width of cloth images under certain yardstick carries out difference, can eliminate the influence of such high-frequency information effectively.

The multispectral alteration that the present invention is based on the wavelet packet variation divides the concrete steps of band method to be:

1. according to the multispectral image that obtains, select suitable wavelet packet basis;

2. with selected wavelet packet basis, respectively the image of two wave bands is carried out the WAVELET PACKET DECOMPOSITION of the suitable number of plies, wherein two wave bands are TM/ETM ⁺Five wave bands (1550～1750nm) and seven wave bands (2080/2090～2350nm);

3. different node coefficients, especially details component coefficient on purpose selecting are 2. found the solution the difference component of two wave band selected node coefficients; Realization is subdued or is suppressed the coefficient of dissociation of non-alteration Information; That is to say, by WAVELET PACKET DECOMPOSITION to 5,7 band images, can form a quad-tree structure, each node on the tree (component or coefficient) restructural one sub-band images, sub-band images by TM5,7 reconstruct has very strong correlation and similarity, and the difference of different images signal has then reflected the fine difference of the spectrum that causes because of alteration to a great extent; Therefore, alteration component (AC:Altered Component) can be expressed as the difference component of different-waveband WAVELET PACKET DECOMPOSITION coefficient (X), promptly

$\mathrm{AC}=\underset{i}{\mathrm{\Σ}}[\underset{j=0}{\overset{2^{2\×i}-1}{\mathrm{\Σ}}}{\left(X_{i,j}\right)}_{\mathrm{ima}{\mathrm{ge}}_1}-\underset{j=0}{\overset{2^{2\×i}}{\mathrm{\Σ}}}{\left(X_{i,j}\right)}_{{\mathrm{image}}_2}]---i=\mathrm{1,2},\·\·\·,n$

4. the difference component sum of node is rebuild the alteration image in utilizing 3.;

5. utilizing 4., the scatter diagram of reconstructed image and arbitrary original image carries out the alteration zone segmentation.

The alteration component reconstruct alteration distribution plan of comprehensive different directions.The alteration image of institute's reconstruct is comprising the abundant alteration hierarchical information of the different numbers of plies, the local different information of transition between especially trickle different alteration zones.Utilize the scatter diagram of reconstructed image and arbitrary original image to segment to alteration zone.

Example:

The Bake beach district exposure of Xinjiang A Erjin zone of fracture has celadon siltstone, sandstone, glutenite and the purple conglomerate folder limestone and the amygdaloidal basalt of Ma Tekebu clarke group, a small amount of Si Laierbu clarke group celadon metasandstone, metamorphic conglomerate, quartzite, pelitic siltstone, crystalline limestone, muddy limestone; Rift structure is grown, and is widely distributed based on the irruptive rock that the late first phase Proterozoic era of ultrabasic rock, diabase and pyroxenolite or graniton etc. is inferior.This district is dry, and physical weathering is strong.Extract Mineralized Alteration mainly is the hydrothermal alteration of tectonization transformation and rock mass invasion in the geology forming process.Data shows that the mineral products in this district mainly originate in the alteration zone of the control intrusive mass of being constructed.As shown in Figure 1, ETM ⁺7, the color composite image of 4,3 wave bands.

" symN " be by Daubechies (Dao Beiqiesi) propose to the improved wavelet basis of tight support biorthogonal wavelet, have near symmetrical and the tight biorthogonal feature that supports.Select " sym3 " wavelet packet basis that this is distinguished ETM ⁺7 and 5 band images decompose, and the difference of reconstruct coefficient of dissociation and image have obtained the alteration Information image (as Fig. 2) in this district.Further utilize the two-dimentional scatter diagram (as Fig. 3) of reconstructed image and TM5 image that the alteration Information of extracting has been carried out segmenting (Fig. 4, Fig. 5).

In mid-term Proterozoic era evening, five magma invasions take place in this district.Invade for the first time for ultrabasic magma or subsiliceous rock slurry, as diabase, graniton and pyroxenolite etc.; Middle partially for the second time acid, lithology has diorite, amphibolite, quartz dioite and granodiorite etc.; Three, be acid magma for the 4th and the 5th time and invade, be porphyritic granite, coarse grain or middle grain biotite granite and moyite etc.In magma invasion and structural evolution process, the rock mass of the invasion in early stage is transformed by later stage invasion rock mass and alteration is taken place, and some stratum horn stoneization etc. occurs as unit in the evening ancient limestone or the sandstone of system down.Contrast geologic map as can be known, the alteration Information that extracts mainly is positioned on ultrabasic rock body, fracture side or the limestone horn stone band of invading for the first time the late mid-term Proterozoic era, is strip and divides band, and supposition should be the reflection of altered rock.

Claims (3)

1. the multispectral alteration based on the wavelet packet variation divides the band method, comprises the steps:

1. according to the multispectral image that obtains, select suitable wavelet packet basis;

2. with selected wavelet packet basis, respectively the image of two wave bands is carried out the WAVELET PACKET DECOMPOSITION of the suitable number of plies;

3. different node coefficients, especially details component coefficient on purpose selecting are 2. found the solution the difference component of two wave band selected node coefficients; Realization is subdued or is suppressed the coefficient of dissociation of non-alteration Information;

4. the difference component sum of node is rebuild the alteration image in utilizing 3.;

5. utilizing 4., the scatter diagram of reconstructed image and arbitrary original image carries out the alteration zone segmentation.

2. method according to claim 1 is characterized in that described two wave band images are TM/ETM ⁺Five wave bands (1550～1750nm) and seven wave bands (2080/2090～2350nm).

3. according to claim 1 or 2 described methods, it is characterized in that alteration component (AC:AlteredComponent) can be expressed as the difference component of different-waveband WAVELET PACKET DECOMPOSITION coefficient (X), the difference component of the WAVELET PACKET DECOMPOSITION coefficient of different-waveband is calculated by following formula:

$\mathrm{AC}=\underset{i}{\mathrm{\Σ}}[\underset{j=0}{\overset{2^{2\×i}-1}{\mathrm{\Σ}}}{\left(x_{i,j}\right)}_{{\mathrm{image}}_1}-\underset{j=0}{\overset{2^{2\×i}}{\mathrm{\Σ}}}{\left(X_{i,j}\right)}_{{\mathrm{image}}_2}]---i=\mathrm{1,2},\·\·\·,n$