A method for detecting aircraft wakes using hyperspectral images based on spatial-spectral joint

Abstract

The present invention discloses a method for detecting aircraft wakes from hyperspectral images based on spatial-spectral integration, comprising: extracting edges band by band in the spatial dimension of a hyperspectral image, statistically merging the generated edge images across all bands, and performing threshold screening on the merged results; performing eigenvalue decomposition on the spectral dimension of the hyperspectral image, screening eigenvectors based on eigenvalues, and reconstructing the image using the retained eigenvectors; calculating pixel anomaly scores of the reconstructed image using an RX operator, and performing threshold screening on the anomaly score results; performing bidirectional verification and merging on the two sets of results obtained from the spatial and spectral dimensions using a bidirectional verification model, and screening the spectral dimension feature detection results using the spatial dimension feature detection results as a guide image, thereby ultimately completing the detection of aircraft wakes. The present invention maximizes the use of the spatial structural information and spectral information of the hyperspectral image, achieving a detection effect with high accuracy and low false alarm rate.

Description

Technical Field

[0001] The present invention relates to the technical field of hyperspectral remote sensing imaging, and in particular to a method for detecting aircraft wakes using hyperspectral images based on space-spectrum combination. Background Art

[0002] Compared with traditional multispectral scanners, imaging spectrometers can obtain continuous images of hundreds of bands, and a spectral curve can be extracted for each image pixel, which has significant advantages for the fine identification of ground objects. The hyperspectral images obtained by imaging spectrometers have the characteristics of "image and spectrum integration", which contain rich spatial, radiation and spectral triple information of ground objects, and can record the characteristics of ground objects in a true and detailed manner. Compared with airborne hyperspectral remote sensing, satellite-borne hyperspectral imaging technology not only maintains a high spectral resolution, but also has the advantages of a large detection range and small limitations. Specifically, a large detection range refers to the larger width of satellite-borne hyperspectral imaging and fast imaging speed. It only takes a few minutes to obtain spectral information of ground objects within hundreds of thousands of square kilometers; small limitations refer to the fact that satellite-borne platforms can cover observation areas around the world that are difficult for airborne platforms to reach.

[0003] The detailed spectral information in hyperspectral images is extremely advantageous for the precise identification of ground objects. It can identify ground objects that are difficult to distinguish using panchromatic and multispectral images, giving it an inherent advantage in target detection tasks. However, the rich ground object information and wide format make the raw hyperspectral image data very large. Moreover, because the three-dimensional data structure of hyperspectral images has physical significance, the spatial correlation between different pixels makes their spectral vectors non-independent, which also poses challenges to existing image processing and data analysis methods.

[0004] Hyperspectral remote sensing target detection is a binary classification problem that separates an image into background and a target of interest. Depending on whether the target's spectral information is known, this problem can be categorized into detection of known target spectral information and anomaly detection of unknown target spectral information. However, in practical applications, the target spectrum is often uncertain due to various factors, such as the environment, lighting, imaging system, and surrounding objects, requiring estimation from the image. Currently, anomaly detection algorithms based on model optimization can be roughly divided into three categories: statistical model-based methods, representation model-based methods, and decomposition model-based methods.

[0005] Anomaly detection algorithms based on statistical models are among the earliest developed methods in anomaly detection. They feature simple models and excellent performance, leveraging the statistical properties of the target and background to detect anomalies. The most classic of these is the Reed-Xiaoli (RX) algorithm proposed by IS Reed and Xiaoli Yu. The RX algorithm is often considered a benchmark algorithm in statistical models. The RX algorithm is a constant false alarm rate (CFAR) algorithm based on a generalized likelihood ratio test. It assumes that the background follows a Gaussian distribution and estimates the probability density function by calculating the covariance and mean vector of all background samples. Detection pixels where anomalies are located will deviate from the primary statistical features represented by the covariance matrix, effectively detecting anomalies. However, since the background in actual remote sensing images often does not follow a single multivariate Gaussian distribution, its detection performance is significantly affected by the background distribution.

[0006] Representation-based methods consider anomaly detection from a single pixel perspective. The basic idea is to select a set of representative background spectra from the image as a background dictionary. These spectra in the background dictionary are also called atoms. Image reconstruction is performed based on the assumption that background pixels can be well represented by the atoms in the background dictionary, while anomaly pixels cannot.

[0007] Methods based on decomposition models focus on describing the overall image features from a global perspective and using this to detect anomalies. Background pixels exhibit a certain degree of spatial and spectral correlation, which manifests as a low-rank characteristic in the data. Anomalous targets have a low similarity to the background, exhibiting a sparse characteristic. However, these two methods either require iterative computations, resulting in high time complexity, or ignore the relationship between spatial and spectral correlations between hyperspectral image pixels, making them ineffective in detecting aircraft wakes. Summary of the Invention

[0008] In response to the above problems, the present invention provides a method for detecting aircraft wakes from hyperspectral images based on spatial-spectral combination, which can break through the background limitation and use the spatial-spectral combination information to detect aircraft wakes. It maximizes the use of the spatial structure information and spectral information of the hyperspectral image to obtain detection results with high accuracy and low false alarm rate. It is highly robust under various complex background distribution conditions and has good detection effect.

[0009] A method for detecting aircraft wakes using hyperspectral images based on spatial-spectral combination includes the following steps:

[0010] (1) Extract the edge of the hyperspectral image band by band in the spatial dimension, statistically merge the generated edge image across all bands, perform threshold screening on the merged result, and obtain the spatial dimension feature detection result;

[0011] (2) Perform eigenvalue decomposition on the spectral dimension of the hyperspectral image, screen the eigenvectors according to the eigenvalue size, and select the retained eigenvectors to reconstruct the image; use the RX operator to calculate the pixel anomaly score of the reconstructed image, perform threshold screening on the anomaly score result, and obtain the spectral dimension feature detection result;

[0012] (3) A bidirectional verification model is used to perform bidirectional verification and merging on the two sets of results obtained from the spatial dimension and the spectral dimension. Based on the idea of ​​guided filtering, a sliding window is used for opposite-side comparison. The side with higher outlier statistical results is defined as the plume, and the other side is defined as the background. The spatial dimension feature detection results are used as a guide image to screen the spectral dimension feature detection results, and finally the detection of the aircraft wake is completed.

[0013] In step (1), the optimal edge detection operator Canny operator is used to extract the edge. The specific process is as follows:

[0014] The characteristic band image is converted into a grayscale image I band by band, and then the grayscale image I is smoothed by Gaussian convolution filtering, and then the gradient is calculated:

[0015]

[0016] Where h is the Gaussian filter, is a differential filter in the spatial dimension x and y directions, the amplitude of the image gradient is G, and the calculated image gradient map is S;

[0017] Subsequently, along the gradient direction, points with non-maximum gradient values ​​are suppressed to achieve edge refinement and single-pixel wide edge detection effects; dual threshold detection and connection are used to ensure a low error rate of the detection results, and a series of edge images of the characteristic bands of the original image are obtained.

[0018] The generated edge image is statistically merged for all bands, and the merged results are threshold filtered. The specific process is as follows:

[0019] After obtaining the edge image, Hough line detection is performed on it. Among the line detection results obtained from the same band image, the Bresenham algorithm is used to merge single-pixel connections for lines with the same parameters to maintain the accuracy of single-pixel wide edge detection. Finally, the detection results of all feature bands are merged using a voting mechanism, and the threshold function f(t) is set to process the detection results:

[0020]

[0021] threshold1=ε1·max(t),0<ε1<1

[0022] Where t represents the number of times the pixel is counted as an edge in the line detection, threshold1 is the threshold number of times it is finally retained as an edge, ε1 is a constant, and f(t) is the statistical result of the corresponding position after screening;

[0023] After the above steps, the spatial dimension line pixel voting result is obtained, that is, the spatial dimension feature detection result.

[0024] In step (2), the spectral dimension of the hyperspectral image is decomposed into eigenvalues ​​using the following formula:

[0025]

[0026] Where S is the hyperspectral image, V is the eigenvalue vector matrix, ∧ is the eigenvalue matrix corresponding to the eigenvector, v1…v N represents the eigenvalue vector, λ1…λ N Indicates the eigenvalue corresponding to the eigenvector.

[0027] The eigenvectors are screened according to the eigenvalues, and the retained eigenvectors are selected to reconstruct the image. Specifically:

[0028] Analyze the eigenvalue histogram, retain the eigenvectors corresponding to the larger eigenvalues, suppress the noise, and reconstruct the image

[0029]

[0030] Where, Represents the reconstructed eigenvector matrix, m is the number of deleted eigenvectors; then the reconstructed image Expressed as:

[0031]

[0032] Where S is the hyperspectral image, μ H is the mean of the image.

[0033] The RX operator is used to calculate the pixel anomaly score of the reconstructed image, and the threshold value is screened for the anomaly score. The specific process is as follows:

[0034] Apply the RX detection operator D to the reconstructed image RX (z) Calculate pixel anomaly score:

[0035]

[0036] Where z is the image pixel, Σ0 is the skew variance matrix, and μ0 is the mean;

[0037] Finally, taking into account the background components in the features retained during feature selection, the results of the anomaly score are thresholded:

[0038] threshold2=ε2max(z),0≤ε2<1

[0039] In the formula, threshold2 is the final retained outlier threshold, ε2 is a constant, and z is the image pixel;

[0040] After the above steps, the anomaly score matrix is ​​obtained, that is, the spectral dimension feature detection result.

[0041] The specific process of step (3) is:

[0042] Based on the assumption that the anomaly score on the plume side is higher than that on the background side, and in accordance with the idea of ​​guided filtering, a sliding window is used, with the center of the window moving along the target position of the spatial dimension feature detection result. During the process, the degree of anomaly of the local range of the corresponding position in the spectral dimension feature detection result is dynamically compared;

[0043] Assume that the final result matrix is ​​Σ, It represents the submatrix divided by the window with the coordinate (i, j) as the center and the odd number k as the dimension. Matrix A is the spatial dimension feature detection result, and matrix B is the spectral dimension feature detection result. Both A and B have the same spatial dimension as the original three-dimensional hyperspectral image. When the element a at position (i, j) in matrix A is i,j When >0, the calculation is performed. The two-way validation model is expressed as follows using matrix operations:

[0044]

[0045] f(X,Y)=H(X·YX·Y T )

[0046] Among them, * operation is the Hadamard product of matrices, which is the result of multiplying the elements of corresponding positions of two matrices of the same size; J k,k is a k-order all-one matrix; I k is the unit matrix; G k is a k-order matrix with 1 only in the center and 0 in the rest of the positions; T k is a k-order upper triangular matrix with all values ​​1. To simplify the operation relationship, the parameters X and Y in the formula represent matrices of the same size. σ is a constant, and the unit step function H(l) is defined as:

[0047]

[0048] Where l represents the result of matrix comparison; after the two-way verification process is completed, the final detection result is obtained.

[0049] In the absence of plume, there is no gradient trend in the outliers on both sides, while in the presence of plume, there must be a gradient trend in the local areas on both sides. The two-way verification improves the positioning accuracy while solving the problem of direction determination on the plume side.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] 1. The present invention fully considers the spatial-spectral correlation of hyperspectral images, uses a spatial-dimensional feature extraction unit improved based on Hough transform to extract important edge information in the image, and effectively reduces the false alarm rate by using spatial-dimensional features.

[0052] 2. The present invention adopts a bidirectional verification model and proposes a guided filtering window, which reduces the amount of calculation while effectively improving the accuracy of the detection results.

[0053] 3) The method proposed in the present invention is not targeted at a certain background distribution and can effectively detect a variety of complex backgrounds. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a flow chart of a method for detecting aircraft wakes using hyperspectral images based on spatial-spectral combination according to an embodiment of the present invention.

[0055] Figure 2 Schematic diagram of the fusion process of the bidirectional verification model in an embodiment of the present invention.

[0056] Figure 3 This is a comparison chart of the detection results of the method of the present invention and other methods. DETAILED DESCRIPTION

[0057] The present invention will be described in further detail below with reference to the accompanying drawings and examples. It should be noted that the following examples are intended to facilitate understanding of the present invention and do not have any limiting effect on the present invention.

[0058] In this embodiment, an image acquired by the visible shortwave infrared hyperspectral camera of the Gaofen-5 satellite is selected as an example, and the image is cropped to 200×200×180.

[0059] like Figure 1 As shown, a method for detecting aircraft wakes based on hyperspectral images of space and spectrum is provided, which includes the following steps:

[0060] Step 1: Extract the edge of the hyperspectral image band by band in the spatial dimension, perform statistical merging of the generated edge image across all bands, perform threshold screening on the merged result, and obtain the spatial dimension feature detection result.

[0061] A cropped hyperspectral image of 200 × 200 × 180 pixels (with 200 rows and columns and 180 bands) was selected for spatial line extraction. Line extraction involves the following steps: The kth (1≤k≤180) spectrum is first converted into a grayscale image. Edge detection is performed using the Canny operator to obtain 180 gradient maps of the image. Hough line detection is performed on each gradient map to obtain 180 edge response maps, or 180 binary matrices of 200 × 200 pixels. These 180 binary matrices are summed and filtered using a threshold of ε1 (0 < ε1 < 1) times the maximum value. The resulting matrix is ​​then filtered to obtain the spatial detection result.

[0062] Among them, the operation process of the Canny operator to calculate the gradient map is as follows:

[0063]

[0064] Where h is the Gaussian filter, is a differential filter in the spatial dimension x and y directions, the amplitude of the image gradient is G, and the calculated image gradient map is S.

[0065] Step 2: Perform eigenvalue decomposition on the spectral dimension of the hyperspectral image, screen the eigenvectors according to the size of the eigenvalues, and select the retained eigenvectors to reconstruct the image; use the RX operator to calculate the pixel anomaly score of the reconstructed image, perform threshold screening on the anomaly score result, and obtain the spectral dimension feature detection result.

[0066] The spectral dimension of the hyperspectral image is regarded as a spectral vector of 200×200 dimensions and 180×1, and eigenvalue decomposition is performed. S is the hyperspectral image, V is the eigenvalue vector matrix, and Λ is the eigenvalue matrix corresponding to the eigenvector:

[0067]

[0068] Get the eigenvalues ​​and eigenvectors, discard the eigenvectors with smaller eigenvalues, and use the remaining eigenvectors to reconstruct the image. m is the number of deleted eigenvectors. The mean value of the image is μ H , then the reconstructed image It can be expressed as:

[0069]

[0070] Take the RX operator D for the reconstructed image RX (z) Perform anomaly detection to obtain an outlier image, and set a threshold for the outlier image to filter it:

[0071]

[0072] threshold2=ε2max(z),0≤ε2<1

[0073] After the above steps, the anomaly score matrix is ​​obtained, that is, the spectral dimension detection result.

[0074] In step 3, a bidirectional verification model is used to perform bidirectional verification and merging on the two sets of results obtained in the spatial dimension and the spectral dimension. The spectral dimension feature detection results are screened using the spatial dimension feature detection results as a guide image, and the detection of aircraft wakes is finally completed.

[0075] Spatial-spectral bidirectional verification. Based on the idea of ​​guided filtering, a sliding window is used, and the center of the window moves along the target position of the spatial dimension feature extraction result. During the process, the abnormality degree of the local range of the corresponding position in the spectral dimension detection result is dynamically compared. The process is simplified as follows Figure 2 As shown. Assume that the final result matrix is ​​Σ, It represents the submatrix divided by the window with odd dimensions (here k = 5, and (i, j) corresponds to the center of the window) with coordinates (i, j) as the center. Matrix A is the boundary response matrix obtained by the spatial dimension feature extraction module, that is, the spatial dimension detection result. Matrix B is the outlier matrix obtained by the spectral dimension feature extraction module, that is, the spectral dimension detection result. Both A and B have the same spatial dimension size as the original three-dimensional hyperspectral image. When the position (i, j) in matrix A is at the center of the window and element a i,j When >0, the calculation is performed. The two-way validation model can be expressed by matrix operations as follows:

[0076]

[0077] f(X,Y)=H(X·YX·Y T )

[0078] Among them, * operation is the Hadamard product of matrices, which is the result of multiplying the elements of corresponding positions of two matrices of the same size; J k,k is a k-order all-one matrix; I k is the unit matrix; G k is a k-order matrix with 1 only in the center and 0 in the rest of the positions; T k is a k-order upper triangular matrix with all values ​​1. To simplify the operation, the parameters X and Y in the formula represent matrices of the same size. σ is a constant, where the unit step function H(l) is defined as:

[0079]

[0080] After the two-way verification process is completed, the final test result is obtained.

[0081] The results of the test by the present invention are compared with those of other methods and the evaluation index table is shown in Table 1 below.

[0082] Table 1

[0083]

[0084] Figure 3 This is a comparison diagram of the detection results of the method of the present invention and other methods. In the figure, (a) is a single-band image, (b) is the labeled true value map, (c) is the experimental result of the RX algorithm, (d) is the experimental result of the LRX algorithm, (d) is the experimental result of the LRASR algorithm, (f) is the experimental result of the KIFD algorithm, and (g) is the experimental result of the method of the present invention.

[0085] According to Table 1 and Figure 3 It can be seen that the detection results of the present invention are closest to the true value graph, the detection result has a high accuracy rate and a low false alarm rate, and the performance of the detection method is excellent.

[0086] The embodiments described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for detecting aircraft wakes based on hyperspectral images using spatial-spectral combination, characterized in that: The following steps are involved: (1) Extract the edge of the hyperspectral image band by band in the spatial dimension, statistically merge the generated edge image across all bands, perform threshold screening on the merged result, and obtain the spatial dimension feature detection result; (2) Perform eigenvalue decomposition on the spectral dimension of the hyperspectral image, screen the eigenvectors according to the eigenvalue size, and select the retained eigenvectors to reconstruct the image; use the RX operator to calculate the pixel anomaly score of the reconstructed image, perform threshold screening on the anomaly score result, and obtain the spectral dimension feature detection result; (3) A bidirectional verification model is used to perform bidirectional verification and merging of the two sets of results obtained from the spatial dimension and the spectral dimension. Based on the idea of ​​guided filtering, a sliding window is used for bilateral comparison. The side with higher abnormal value statistical results is defined as plume, and the other side is defined as background. The spatial dimension feature detection results are used as a guide image to screen the spectral dimension feature detection results, and finally the detection of aircraft wakes is completed. The specific process is as follows: Based on the assumption that the anomaly score on the plume side is higher than that on the background side, and in accordance with the idea of ​​guided filtering, a sliding window is used, with the center of the window moving along the target position of the spatial dimension feature detection result. During the process, the degree of anomaly of the local range of the corresponding position in the spectral dimension feature detection result is dynamically compared; Assume that the final result matrix is ​​Σ, It represents the submatrix divided by the window with the coordinate (i, j) as the center and the odd number k as the dimension. Matrix A is the spatial dimension feature detection result, and matrix B is the spectral dimension feature detection result. Both A and B have the same spatial dimension as the original three-dimensional hyperspectral image. When the element a at position (i, j) in matrix A is i,j When >0, the calculation is performed. The two-way validation model is expressed as follows using matrix operations: f(X,Y)=H(X·Y-X·Y T ) Among them, * operation is the Hadamard product of matrices, which is the result of multiplying the elements of corresponding positions of two matrices of the same size; J k,k is a k-order all-one matrix; I k is the unit matrix; G k is a k-order matrix with 1 only in the center and 0 in the rest of the positions; T k is a k-order upper triangular matrix with all values ​​1. To simplify the operation relationship, the parameters X and Y in the formula represent matrices of the same size. σ is a constant, and the unit step function H(l) is defined as: Where l represents the result of matrix comparison; after the two-way verification process is completed, the final detection result is obtained.

2. The method for detecting aircraft wakes based on hyperspectral imagery using spatial-spectral combined methods according to claim 1, characterized in that: In step (1), the optimal edge detection operator Canny operator is used to extract the edge. The specific process is as follows: The characteristic band image is converted into a grayscale image I band by band, and then the grayscale image I is smoothed by Gaussian convolution filtering, and then the gradient is calculated: Where h is the Gaussian filter, is a differential filter in the spatial dimension x and y directions, the amplitude of the image gradient is G, and the calculated image gradient map is S; Subsequently, along the gradient direction, points with non-maximum gradient values ​​are suppressed to achieve edge refinement and single-pixel wide edge detection effects; dual threshold detection and connection are used to ensure a low error rate of the detection results, and a series of edge images of the characteristic bands of the original image are obtained.

3. The method for detecting aircraft wakes based on hyperspectral imagery using spatial-spectral combination according to claim 1, characterized in that: In step (1), the generated edge image is statistically merged for all bands, and the merged result is threshold-filtered. The specific process is as follows: After obtaining the edge image, Hough line detection is performed on it. Among the line detection results obtained from the same band image, the Bresenham algorithm is used to merge single-pixel connections for lines with the same parameters to maintain the accuracy of single-pixel wide edge detection. Finally, the detection results of all feature bands are merged using a voting mechanism, and the threshold function f(t) is set to process the detection results: threshold1=ε1·max(t),0<ε1<1 Where t represents the number of times the pixel is counted as an edge in the line detection, threshold1 is the threshold number of times it is finally retained as an edge, ε1 is a constant, and f(t) is the statistical result of the corresponding position after screening; After the above steps, the spatial dimension line pixel voting result is obtained, that is, the spatial dimension feature detection result.

4. The method for detecting aircraft wakes based on hyperspectral imagery using spatial-spectral combination according to claim 1, wherein: In step (2), the spectral dimension of the hyperspectral image is decomposed into eigenvalues ​​using the following formula: Where S is the hyperspectral image, V is the eigenvalue vector matrix, Λ is the eigenvalue matrix corresponding to the eigenvector, v1…v N represents the eigenvalue vector, λ1…λ N Indicates the eigenvalue corresponding to the eigenvector.

5. The method for detecting aircraft wakes based on hyperspectral imagery using spatial-spectral combination according to claim 4, characterized in that: In step (2), the eigenvectors are screened according to the eigenvalues, and the retained eigenvectors are selected to reconstruct the image, specifically: Analyze the eigenvalue histogram, retain the eigenvectors corresponding to the larger eigenvalues, suppress the noise, and reconstruct the image Where, Represents the reconstructed eigenvector matrix, m is the number of deleted eigenvectors; then the reconstructed image Expressed as: Where S is the hyperspectral image, μ H is the mean of the image.

6. The method for detecting aircraft wakes based on hyperspectral imagery using spatial-spectral combination according to claim 1, characterized in that: In step (2), the RX operator is used to calculate the pixel anomaly score of the reconstructed image, and the result of the anomaly score is threshold-screened. The specific process is as follows: Apply the RX detection operator D to the reconstructed image RX (z) Calculate pixel anomaly score: Where z is the image pixel, Σ0 is the skew variance matrix, and μ0 is the mean; Finally, taking into account the background components in the features retained during feature selection, the results of the anomaly score are thresholded: threshold2=ε2max(z),0≤ε2<1 In the formula, threshold2 is the final retained outlier threshold, ε2 is a constant, and z is the image pixel; After the above steps, the anomaly score matrix is ​​obtained, that is, the spectral dimension feature detection result.

7. The method for detecting aircraft wakes based on hyperspectral imagery using spatial-spectral combination according to claim 1, characterized in that: In the absence of plume, there is no gradient trend in the outliers on both sides, while in the presence of plume, there must be a gradient trend in the local areas on both sides. The two-way verification improves the positioning accuracy while solving the problem of direction determination on the plume side.

Images