CN110119716A - A kind of multi-source image processing method - Google Patents

Abstract

An embodiment of the present invention provides a multi-source image processing method, including: adopting automatic clustering to obtain first clustering labels of multiple related images; wherein, the multiple related images include multiple images and An image related to a location or a target; extracting invariant features and differential features of the plurality of related images based at least on the basis of the first clustering label; and performing image post-processing according to the invariant features and the differential features , wherein the image post-processing includes target recognition or image fusion. The invention comprehensively utilizes multi-source remote sensing images, extracts and interprets invariant features and differential features of different levels and scales of different sensor images from the data itself without prior knowledge. The invention can be widely used in multi-source remote sensing image fusion and target recognition.

Description

A Multi-source Image Processing Method

technical field

The invention relates to the technical fields of multi-source remote sensing image fusion, feature understanding, expert interpretation, etc., and specifically relates to a multi-source image processing method.

Background technique

The key to feature interpretation of multi-source remote sensing images is to extract common invariant features and differential complementary features of multi-source remote sensing images. Due to the differences in imaging mechanisms, it is very difficult to automatically extract the common invariant features and differential complementary features of multi-source remote sensing images, which seriously restricts the practical application of multi-source heterogeneous remote sensing images.

Contents of the invention

In order to solve the above technical problems, the present invention proposes a multi-source image processing method.

The present invention proposes a multi-source image processing method, including: adopting automatic clustering to obtain a first clustering label of a plurality of related images; wherein, the plurality of related images include a plurality of images obtained by using a plurality of sensors and locations or Target-related images; extracting invariant features and differential features of the plurality of related images based at least on the basis of the first clustering class; and performing image post-processing according to the invariant features and the differential features, wherein , the image post-processing includes target recognition or image fusion.

In some embodiments, the multi-source image processing method further includes: generating a composite image based on the multiple related images, and using automatic clustering to obtain the second cluster class label and multi-source invariant samples of the composite image set, wherein the second clustering class mark includes the clustering center and the category of each pixel; based on the multiple related images, automatic clustering is used to obtain a multi-source difference sample set, and according to the second clustering class resetting the first clustering class mark; and based on the multi-source difference sample set, the multi-source invariance sample set, the first clustering class mark, the second clustering class mark, and The feature explanation equation extracts the invariant feature and the difference feature of the plurality of related images.

In some embodiments, the multi-source image processing method further includes extracting a residual image based on the multiple related images and the invariant features, performing multi-scale and multi-level feature interpretation on the residual image, extracting Invariance and complementarity features of the residual image.

In some embodiments, the multi-scale and multi-level feature interpretation is an iterative process of extracting invariant features and complementary features on the residual multi-source images of the current level.

In some embodiments, the multiple related images are multi-source remote sensing images obtained by using multiple sensors.

In some embodiments, the integrated images are automatically clustered based on similarities and differences of common features of the multi-source remote sensing images.

In some embodiments, the multi-source remote sensing images include multiple single-source images; the multi-source remote sensing images are automatically clustered based on the similarities and differences of the single-source images.

In some embodiments, the clustering labels of the single-source image are reset based on the cluster center and the label of the integrated image.

In some embodiments, the characteristic interpretation equation is constructed based on a multi-source invariant characteristic Laplacian matrix, a multi-source complementary characteristic Laplacian matrix and an order-preserving characteristic Laplacian matrix.

In some embodiments, the multi-source image processing method further includes: constructing a multi-source invariance feature based on the multi-source invariant sample set and its class label, the multi-source difference sample set and its class label matrix, multi-source complementary characteristic matrix and order-preserving characteristic matrix; solving the characteristic interpretation equation to obtain the generalized eigenvalue and the generalized eigenvector; the generalized eigenvector corresponding to the non-zero minimum generalized eigenvalue of the characteristic interpretation equation generating a feature interpretation projection matrix; and extracting invariant features and difference features of the plurality of related images based on the feature interpretation projection matrix.

The method of the present invention has important significance for interpreting the feature invariance and difference of multi-source remote sensing images, understanding and solving the difficulties of multi-source data fusion and target recognition, and its main advantages are as follows:

The present invention considers the complexity and versatility of multi-source remote sensing image feature interpretation, uses the information of multi-source image itself to mine feature invariance and difference, does not need training samples and prior knowledge, has no limit to the number of image types, and has a wide range of universality.

The invention fully considers the dependence of feature invariance and difference on the category and scale of ground objects, constructs a feature interpretation equation based on the similarity and difference of clustering objects through multi-source remote sensing images, and gradually and iteratively extracts different levels, Feature invariance and variability across different scales greatly improves feature interpretation capability and performance.

Benefiting from the above advantages, the present invention greatly improves the performance of multi-source remote sensing image interpretation, provides good support for multi-source remote sensing image fusion and feature interpretation, and can be widely used in multi-source remote sensing image fusion, target recognition, scene classification system.

Description of drawings

Fig. 1 is a multi-source image processing method provided by an embodiment of the present invention.

Fig. 2 is a multi-source image processing method provided by an embodiment of the present invention.

Fig. 3 is a flow chart of multi-source remote sensing image feature interpretation provided by another embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings. Referring to the non-limiting exemplary embodiments shown in the accompanying drawings and detailed in the following description, the examples of the present disclosure will be more fully described. Embodiments and their various features and advantageous details. It should be noted that the features shown in the figures are not necessarily drawn to scale. This disclosure omits descriptions of well-known components and techniques so as not to obscure the example embodiments of the disclosure. The examples given are only intended to facilitate understanding of the implementation of the example embodiments of the present disclosure and to further enable those skilled in the art to practice the example embodiments. Accordingly, these examples should not be construed as limiting the scope of embodiments of the present disclosure.

Unless otherwise specifically defined, the technical terms or scientific terms used in the present disclosure shall have the usual meanings understood by those skilled in the art to which the present disclosure belongs. "First", "second" and similar words used in this disclosure do not indicate any order, quantity or importance, but are only used to distinguish images of different features. In addition, in the various embodiments of the present disclosure, the same or similar reference numerals denote the same or similar components.

There is a strong complementarity between multi-source heterogeneous remote sensing images. The comprehensive utilization of the complementarity of multi-source heterogeneous remote sensing images plays an important role in improving the performance of target recognition. Multi-source remote sensing image feature interpretation plays an important role in multi-source remote sensing image fusion, There are urgent application requirements in many application fields such as target recognition and feature understanding.

The embodiment of the present invention starts from the multi-source remote sensing image itself, and automatically extracts the invariant and complementary features of the multi-source remote sensing image without relying on training samples or expert knowledge. The application provides a reliable basis for decision-making.

The embodiment of the present invention provides an effective method for interpreting features of multi-source remote sensing images in view of difficulties in interpreting features of multi-source remote sensing images and requirements of practical applications.

As shown in FIG. 1 , the present invention provides a multi-source image processing method. The image processing method includes: step 101, adopting automatic clustering to acquire first cluster labels of multiple related images; wherein, the multiple related images include multiple images related to places or objects acquired by multiple sensors ; Step 102, extracting invariant features and difference features of the plurality of related images based at least on the basis of the first clustering class label; and Step 103, performing image post-processing according to the invariance features and the difference features , wherein the image post-processing includes target recognition or image fusion. For example, the multiple related images involved in step 101 may be multi-source remote sensing images obtained by using multiple sensors. In some examples, the integrated images are automatically clustered based on similarities and differences of common features of the multi-source remote sensing images.

Using automatic clustering to process multiple related images can start from multiple related images (for example, multi-source remote sensing images) and automatically extract multiple related images (multi-source remote sensing images) without relying on training samples or expert knowledge. )'s invariant and complementary features can provide reliable decision-making basis for practical applications such as multi-source remote sensing image fusion and target recognition.

As shown in Figure 2, in some embodiments, the multi-source image processing method further includes: step 201, generating a composite image based on the multiple related images, and using automatic clustering to obtain the second clustering class label of the composite image And a multi-source invariant sample set, wherein, the second clustering class mark includes a cluster center and the category of each pixel; Step 101 also includes using automatic clustering to obtain multi-source difference samples based on the multiple related images set, reset the first clustering class according to the second clustering class; and step 102 exemplary includes based on the multi-source difference sample set, the multi-source invariance sample set, the first A clustering label, the second clustering label, and a feature interpretation equation extract the invariant features and differential features of the multiple related images.

In some embodiments, the multi-source image processing method further includes extracting a residual image based on the multiple related images and the invariant features, performing multi-scale and multi-level feature interpretation on the residual image, and extracting the residual Image invariance and complementarity features. For example, the multi-scale and multi-level feature interpretation may include an iterative process of extracting invariant features and complementary features from the residual multi-source images of the current level.

In some embodiments, the multi-source remote sensing images include multiple single-source images; the multi-source remote sensing images are automatically clustered based on the similarities and differences of the single-source images. For example, in some examples, the clustering labels of the single-source image are reset based on the cluster center and the label of the integrated image.

In some embodiments, the characteristic interpretation equation is constructed based on a multi-source invariant characteristic Laplacian matrix, a multi-source complementary characteristic Laplacian matrix and an order-preserving characteristic Laplacian matrix. For example, the multi-source image processing method includes: constructing a multi-source invariant feature matrix, a multi-source complementary feature matrix and An order-preserving eigenmatrix; solving the eigeninterpretation equation to obtain a generalized eigenvalue and a generalized eigenvector; generating an eigeninterpretation projection matrix from the generalized eigenvector corresponding to the non-zero minimum generalized eigenvalue of the eigeninterpretation equation; and based on The feature interpretation projection matrix extracts invariant features and differential features of the multiple related images.

The present invention also provides an embodiment of multi-source remote sensing image processing, as shown in FIG. 3 . The multi-source remote sensing image processing method of the present invention comprises the following steps:

Step S1 generates a composite image from the multi-source remote sensing image, and obtains cluster labels and multi-source invariant samples through automatic clustering on the composite image.

Step S2 obtains the cluster labels and multi-source difference samples through automatic clustering on the single-source image respectively, and resets the cluster labels of the single-source image based on the cluster centers and labels of the integrated image.

Step S3 constructs multi-source invariant feature matrix, multi-source complementary feature matrix and order-preserving feature matrix based on multi-source invariant samples, multi-source differential samples and their class labels, and solves feature interpretation equations to obtain generalized eigenvalues and generalized Eigenvectors and feature interpretation projection matrix are constructed to extract invariant features and difference features.

Step S4 constructs a residual multi-source image based on the multi-source remote sensing image and the invariant feature image, and iterates steps S2 and S3 on the residual multi-source image to extract invariant features and differential features at different levels and scales until the iteration termination is satisfied condition.

Referring to FIG. 3 , the present invention also provides the following example for illustration of the above steps.

Step S1 generates a composite image from the multi-source remote sensing image, and obtains cluster labels and multi-source invariant samples through automatic clustering on the composite image. The specific process of integrated images and automatic clustering is as follows.

Step S11 is to generate a comprehensive image. Let I _k represent the multi-source remote sensing images of the same scene captured by different sensors, and the multi-source remote sensing images have been registered, where K is the number of types of multi-source remote sensing images. The imaging mechanisms of multi-source remote sensing images are different, the characteristics are very different, and the number of bands is also different. For example, I ₁ is a panchromatic image, I ₂ is a multispectral image, I ₃ is an infrared image, I ₄ is a SAR image, and I ₅ is a hyperspectral image. In order to avoid the category imbalance caused by the difference in the number of bands, multi-band images are averaged in the band dimension to obtain grayscale images. Specifically, if the number of bands of the multi-source remote sensing image I _k is 1, then the average grayscale image corresponding to the multi-source remote sensing image I _k is G _k =I _k ; if the number of bands of the multi-source remote sensing image I _k is greater than 1, then The average grayscale image corresponding to the multi-source remote sensing image I _k is G _k =mean(3,I _k ), where mean(3,I _k ) represents the average operation of the multi-source remote sensing image I _k in the band dimension. The comprehensive image X is formed by linking single-band images or average grayscale images in the band dimension, that is, X=cat(3,G ₁ ,G ₂ ,…,G _K ), and cat(3,G ₁ ,G ₂ , ..., G _K ) means to link the grayscale images G ₁ , G ₂ , ..., G _K in the band dimension.

Step S12 automatically clusters. This example is an unsupervised method for feature interpretation of multi-source remote sensing images, with no prior knowledge or training samples available. To this end, this example extracts the similarities and differences of multi-source images by automatically clustering on the synthetic image X. Automatic clustering takes the correlation coefficient of the multi-band gray feature vectors of each pixel of the integrated image X as a similarity measure, and regards all pixels as potential cluster center points during initialization, and then iteratively propagates responsibility information and usability information to find cluster centers for each class and determine the grouping class for each pixel. The sense of responsibility message r(i,k) indicates the degree to which the k-th pixel is suitable as the cluster center of the i-th pixel. The availability message a(i,k) represents the probability that the i-th pixel selects the k-th pixel as its cluster center. The specific process of automatic clustering is as follows:

S121 initial state: availability message a(i,k)=0;

S122 updates all sense of responsibility messages according to availability messages, namely

Among them, s(i,k) represents the cross-correlation coefficient between the multi-band grayscale feature vectors between the i-th pixel and the k-th pixel.

S123 Update all availability messages according to the sense of responsibility message, namely

S124 Combining the availability message and the sense of responsibility message to determine the cluster center. For the i-th pixel, if k is equal to i when the sum of the availability message and the sense of responsibility message, that is, "a(i,k)+r(i,k)", is equal to i, then the pixel i itself is the cluster center; If k is not equal to i, it means that pixel i is a subsidiary point, and its cluster center is pixel k.

S125 If the set maximum number of iterations T is reached or the amount of information change in the data point is less than a given threshold τ, the algorithm ends; otherwise, go to step S122. In this example, the number of iterations T=100, and the threshold τ=10. Those skilled in the art can set the specific trend of the number of iterations T and the specific value of the threshold τ according to actual needs.

S13 At the end of the iteration of the automatic clustering algorithm, the clustering center c _l (l=1, . . . , L) of the integrated image X and the category of each pixel can be obtained. For the convenience of description, the above clustering process is denoted as f ₀ . For any pixel p of the integrated image X, f ₀ (p) represents the cluster label corresponding to the pixel p. The set of multi-band grayscale feature vectors corresponding to pixels of the same category in the integrated image X is called the multi-source invariant sample set of this class. If there are L clustering categories, there are L multi-source invariant sample sets.

Step S2 obtains the cluster labels and multi-source difference samples through automatic clustering on the single-source image respectively, and resets the cluster labels of the single-source image based on the cluster centers and labels of the integrated image. The specific process of the single-source image automatic clustering and class label reset is as follows:

Step S21 Automatic clustering of single-source images. Respectively use the single-source image I _k to replace the comprehensive image X and use the method described in step S12 to automatically cluster the single-source image I _k to obtain the single-source clusterer f _k and the clustering center of each type of image and the category of each pixel. Similarly, the class label samples of the single-source image I _k can be obtained according to the method described in step S13. For convenience of description, use Indicates the i-th sample on the comprehensive image X and the single-source image I _k , and its class label is denoted by Indicates that k=0,…,K.

Step S22 class label reset. Since the clustering processes of integrated images and multi-source images are carried out independently, their class labels lack correspondence. For this reason, the clustering labels of the single-source image I _k are reset based on the clustering centers and labels of the integrated image X. The specific method is: reset the class label of the pixel p of the image I _k to:

Among them, s(c _k (p),c ₀ )) indicates whether the cluster center c _k (p) of pixel p in image I _k has other cluster centers on image I _k and corresponds to the The same cluster center c ₀ (c _k (p)).

Step S3 constructs multi-source invariant feature matrix, multi-source complementary feature matrix and order-preserving feature matrix based on multi-source invariant samples, multi-source differential samples and their class labels, and solves feature interpretation equations to obtain generalized eigenvalues and generalized Eigenvectors and feature interpretation projection matrix are constructed to extract invariant features and difference features. The feature matrix mainly includes multi-source invariant feature matrix, multi-source complementary feature matrix and order-preserving feature matrix. Each type of eigenmatrix includes a fundamental matrix, a row and diagonal matrix, and a Laplacian matrix. The specific process of constructing the feature matrix and feature interpretation equation is as follows:

Step S31: Multi-source invariant feature matrix construction. The basic matrix W _s of multi-source invariance is composed of block matrices, and the specific form is:

It is a matrix with m _a row and m _b column, where m _a and m _b represent the number of samples in I _a and I _b respectively. Indicates the cluster center corresponding to pixel i on the integrated image. condition The meaning of is that the pixel i on the image I _a and the pixel j on the image I _b are of different categories, but their corresponding categories in the integrated image X are the same. Multisource Invariant Row and Diagonal Matrices Multi-source invariant Laplacian matrix L _s =D _s -W _s . for A matrix corresponding to a=0, b=0; for A matrix corresponding to a=0, b=K; for A matrix corresponding to a=K, b=0; for A matrix corresponding to a=K, b=K; is the category of pixel i on image I _a ; is the category of pixel j on image I _b ; is the category corresponding to the pixel i on the image I _a in the integrated image X; is the category corresponding to the pixel j on the image I _b in the integrated image X.

Step S32: Multi-source complementarity feature matrix construction. The multi-source complementarity feature matrix W _d is also composed of block matrices, the specific form is:

Multi-source Complementarity Row and Diagonal Eigen Matrices Multi-source complementarity Laplacian matrix L _d =D _d -W _d . for A matrix corresponding to a=0, b=0; for A matrix corresponding to a=0, b=K; for A matrix corresponding to a=K, b=K; for Corresponding to a=K, b=K matrix.

Step S33: Construction of an order-preserving feature matrix. First construct the single-source order-preserving similarity matrix W _k , the single-source order-preserving row and diagonal feature matrix D _k , and the single-source order-preserving Laplacian L _k , the single-source order-preserving similarity matrix W Each element of _k , single-source order-preserving row and diagonal eigenmatrix D _k , and single-source order-preserving Laplacian L _k is calculated as follows: L _k =D _k -W _k .

The overall topological relationship matrix is

The overall feature matrix is

Wherein, I ₀ =X, and X is the comprehensive image described in step S11 above.

Step S34: Construction of feature interpretation equations and multi-source feature interpretation. Feature interpretation is based on the multi-source invariant Laplacian matrix L _s , the multi-source complementary Laplacian matrix L _d , the overall topological relationship matrix L, and the overall feature matrix Z for feature extraction and feature interpretation. The interpretation equation is: Z(L+L _s )Z ^T μ=λZL _d Z ^T μ. Solve the eigeninterpretation equation to obtain the generalized eigenvalue λ _k and the generalized eigenvector μ _k . The generalized eigenvector μ _k corresponding to the smallest (K+1) non-zero generalized eigenvalues λ _k is the feature projection direction. Specifically, for the feature vector x _i at the i-th pixel in I _k , its expression in I _j is represents the pseudo-inverse of μ _j . The complementarity with I _j is characterized by I is the identity matrix. The common features of _xi in different types of images are μ ₀ represents the generalized eigenvector corresponding to the largest λ _k among (K+1) minimum non-zero generalized eigenvalues.

Step S4 constructs a residual multi-source image based on the multi-source remote sensing image and the invariant feature image, and iterates steps S2 and S3 on the residual multi-source image to extract invariant features and differential features at different levels and scales until the iteration termination is satisfied condition. The specific process of the residual multi-source image extraction and fine feature interpretation is as follows:

Step S41 Residual multi-source image extraction. In order to understand the commonality and complementarity of multi-source remote sensing images at a finer scale, the residual features at the previous scale are calculated first. Specifically, for the feature vector x _i corresponding to each pixel of I _k , calculate the residual feature vector μ ₀ represents the generalized eigenvector corresponding to λ _k with the largest value among (K+1) smallest non-zero generalized eigenvalues, and then the residual eigenvector x _i ' of different pixels is composed of a new image I according to the original row and column position of _xi ' _k .

Step S42 fine feature interpretation. Then, according to the method described in step S2 to step S3, extract multi-source difference samples, construct a feature matrix, and extract invariant features and complementary features on the new multi-source image _I'k .

Step S43 scale switching. If the ratio of the square of each pixel feature vector of I' _k to the sum of the squares of each pixel feature vector of I _k is less than ε, the iteration is terminated; otherwise, scale switching and feature interpretation are performed until the iteration termination condition is met, and the scale switching The specific method of feature interpretation is to update the residual multi-source image I' _k according to the method described in step S41, and perform fine feature interpretation according to the method described in step S42.

The advantage of multi-level feature interpretation is to gradually understand the invariance and differences of multi-source remote sensing images at different levels and scales through the "magnifying glass". In this way, the common invariant features and complementary differential features of multi-source remote sensing images are extracted completely and comprehensively, which is very important for subsequent applications such as feature fusion and target recognition.

To sum up, the embodiment of the present invention provides a multi-source remote sensing image feature interpretation method, including: Step S1: generate a comprehensive image from the multi-source remote sensing image, and obtain cluster labels and multi-source invariance through automatic clustering on the comprehensive image sample. Step S2: Obtain cluster labels and multi-source difference samples through automatic clustering on the single-source image, and reset the cluster labels of the single-source image based on the cluster centers and labels of the integrated image . Step S3: Construct multi-source invariant feature matrix, multi-source complementary feature matrix and sequence-preserving feature matrix based on multi-source invariant samples, multi-source differential samples and their class labels, and solve feature interpretation equations to obtain generalized eigenvalues and Generalized eigenvectors and construct feature interpretation projection matrix to extract invariant features and difference features. Step S4: Construct a residual multi-source image based on the multi-source remote sensing image and the invariant feature image, and iterate step S2 and step S3 on the residual multi-source image to extract invariant features and difference features of different levels and scales until the iteration is satisfied Termination condition. The invention comprehensively utilizes multi-source remote sensing images, extracts and interprets invariant features and differential features of different levels and scales of different sensor images from the data itself without prior knowledge. The invention can be widely used in multi-source remote sensing image fusion and target recognition.

Those skilled in the art should be able to realize that the modules and method steps described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two, and that the programs corresponding to the software modules and method steps Can be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or known in the technical field any other form of storage medium. In order to clearly illustrate the interchangeability of electronic hardware and software, the composition and steps of each example have been generally described in terms of functions in the above description. Whether these functions are performed by electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may implement the described functionality using different methods for each particular application, but such implementation should not be considered as exceeding the scope of the present invention.

The term "comprising" or any other similar term is intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus/apparatus comprising a set of elements includes not only those elements but also other elements not expressly listed, or Also included are elements inherent in these processes, methods, articles, or devices/devices.

So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings, but those skilled in the art will easily understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to relevant technical features, and the technical solutions after these changes or substitutions will all fall within the protection scope of the present invention.

Claims (10)

1. a kind of multi-source image processing method, comprising:

The first cluster category of several associated pictures is obtained using automatic cluster；Wherein, several described associated pictures include using Several images relevant to place or target that multiple sensors obtain；

The Invariance feature and otherness feature of several associated pictures are at least extracted based on the first cluster category；And

Post processing of image is carried out according to the Invariance feature and the otherness feature, wherein described image post-processing packet Include target identification or image co-registration.

2. multi-source image processing method as described in claim 1, further includes:

Synthetic image is generated based on several described associated pictures, the second cluster class of the synthetic image is obtained using automatic cluster It is marked with and multi-source invariance sample set, wherein the second cluster category includes the classification of cluster centre and each pixel；

Multi-source otherness sample set is obtained using automatic cluster based on several described associated pictures, according to the second cluster class Indicated weight sets the first cluster category；And

Based on the multi-source otherness sample set, the multi-source invariance sample set, the first cluster category, described the Two cluster categories and feature explain equation, extract the Invariance feature and otherness feature of several associated pictures.

3. multi-source image processing method as claimed in claim 2, further includes, based on several described associated pictures and it is described not Vertic features extract residual image, and carry out multiple dimensioned multi-level features interpretation to the residual image, extract the remaining figure The Invariance feature of picture and complementary feature.

4. multi-source image processing method according to claim 3, wherein the multiple dimensioned multi-level features interpretation is to work as The iterative process of Invariance feature and complementary feature is extracted on the remaining multi-source image of preceding level.

5. multi-source image processing method as claimed in claim 1,2 or 3, wherein several described associated pictures are using multiple The multi-source Remote Sensing Images that sensor obtains.

6. multi-source image processing method as claimed in claim 5, wherein the common trait based on the multi-source Remote Sensing Images Similitude and otherness carry out automatic cluster to the synthetic image.

7. multi-source image processing method as claimed in claim 6, wherein the multi-source Remote Sensing Images include several Dan Yuantu Picture；

Similitude and otherness based on single source images carry out automatic cluster to the multi-source Remote Sensing Images.

8. multi-source image processing method as claimed in claim 7, wherein be designated as with the cluster centre and class of the synthetic image Benchmark resets the cluster category of single source images.

9. multi-source image processing method as claimed in claim 7, wherein it is special that the feature interpretation equation is based on multi-source invariance Levy Laplacian Matrix, multi-source complementarity feature Laplacian Matrix and isotonicity feature Laplacian Matrix construction.

10. multi-source image processing method as claimed in claim 9, further includes,

Multi-source is constructed based on the multi-source invariance sample set and its category, the multi-source otherness sample set and its category Invariance feature matrix, multi-source complementarity eigenmatrix and isotonicity eigenmatrix；

It solves the feature interpretation equation and obtains generalized eigenvalue and generalized eigenvector；

Feature interpretation projection is generated by the corresponding generalized eigenvector of non-zero minimum generalized eigenvalue of feature interpretation equation Matrix；And

The Invariance feature and otherness feature of several associated pictures are extracted based on feature interpretation projection matrix.