A method for transforming complex image data of polarimetric synthetic aperture radar

Abstract

The present invention belongs to the technical field of image interpretation, and specifically discloses a polarization synthetic aperture radar complex image data transformation method, including: step 1, selecting a SAR complex image; step 2, converting the SAR complex image into an amplitude image A and a phase image φ; step 3, calculating the mean square error σ of the amplitude image A; step 4, calculating the U image according to the mean square error σ of the amplitude image A; step 5, calculating the phase image φ; step 6, calculating the phase V image according to the phase image φ; step 7, the calculation result (U, V) is the final SAR complex image data transformation result. The U and V obtained by the present invention through the transformation method have almost the same numerical range and statistical distribution form, which is conducive to improving the application degree and depth of phase data in SAR image target enhancement and detection, and improving the interpretation accuracy of SAR remote sensing images.

Description

Technical Field

[0001] The invention belongs to the technical field of image interpretation, and in particular relates to a polarization synthetic aperture radar complex image data transformation method. Background Art

[0002] In the process of target enhancement and detection of general synthetic aperture radar (SAR) images, the amplitude image is usually modeled and analyzed, and the CFAR method is used for calculation; while in the target polarization transformation, complex numbers are generally still used for calculation, and the utilization of phase information is relatively low.

[0003] Unlike the case where the pixel values ​​of general optical images are real numbers, the pixel values ​​in synthetic aperture radar (SAR) images are generally complex numbers (I, Q) expressed in algebraic form, representing the coherent superposition of the backscatter of the target echo in the resolution unit of the surface illumination area, which not only includes the backscatter characteristics of the target in the illumination area, but also includes the spatial characteristics between different targets in the pixel content. Through exponential form transformation, the complex value of each pixel in the SAR image can be expressed in amplitude and phase. Generally, the amplitude of polarization SAR images conforms to Rayleigh distribution, while the phase conforms to circular periodic uniform distribution. The numerical range and statistical distribution function form of the two are completely different.

[0004] At present, the data transformation methods of single-polarization SAR images mainly include intensity transformation of image pixel amplitude and logarithmic transformation of image pixel amplitude. In the original SAR complex image, the phase follows a uniform distribution between -π and π, the amplitude follows a Rayleigh distribution, and the intensity follows an exponential distribution. Whether it is an amplitude image or an intensity image, the distribution of the pixel value range varies greatly, and the range of variation of SAR complex images in different regions may be different, and the difference is large. The logarithmic transformation introduces negative numbers in the image pixel value. For targets with smaller echoes, the absolute value of the logarithmic value is relatively large, which is easy to be confused with targets with larger echo intensity, resulting in a loss of target accuracy.

[0005] The present invention aims to propose a data transformation method to transform SAR complex image data from amplitude and phase space to a new two-dimensional data space, namely (I,Q)→(U,V), under which U and V have almost the same value range and statistical distribution form. Summary of the invention

[0006] The purpose of the present invention is to provide a polarimetric synthetic aperture radar complex image data transformation method, wherein U and V obtained by the transformation have almost the same numerical range and statistical distribution form, which is beneficial to improving the application degree and depth of phase data in SAR image target enhancement and detection, and improving the interpretation accuracy of SAR remote sensing images.

[0007] The technical solution to achieve the purpose of the present invention is:

[0008] A method for transforming complex image data of a polarization synthetic aperture radar, the method comprising the following steps:

[0009] Step 1, select SAR complex image;

[0010] Step 2: Convert the SAR complex image into an amplitude image A and a phase image φ;

[0011] Step 3, calculate the mean square error σ of the amplitude image A;

[0012] Step 4: Calculate the U image according to the mean square error σ of the amplitude image A;

[0013] Step 5, calculate the phase image φ;

[0014] Step 6, calculating the phase V image according to the phase image φ;

[0015] Step 7: The calculation result (U, V) is the final SAR complex image data transformation result.

[0016] The step 1 is specifically as follows: select the entire SAR complex image or a sub-image containing some pixels, and assume that the number of rows is M and the number of columns is N, then the total number of pixels in the image is M×N, and the subscript (i, j) represents that the specific position of the pixel in the image is the i-th row and the j-th column.

[0017] The step 2 specifically includes: using the Euler formula to convert the pixel complex number from an algebraic form to an exponential form of a complex number, thereby forming an amplitude image A and a phase image φ.

[0018] The formula for calculating the mean square error σ of the amplitude image A in step 3 is:

[0019]

[0020] Where σ is the mean square error of the amplitude image, A i,j is the pixel amplitude value of the i-th row and j-th column, A m is the mean of image A, M is the total number of rows of amplitude image A, and N is the total number of columns of amplitude image A.

[0021] The step 4 is specifically as follows: for the amplitude image A, pixel by pixel calculation The value of is assigned accordingly to form a new image and recorded as U, that is,

[0022] The step 5 is specifically as follows: for the phase image φ, calculate pixel by pixel, if φ i,j <0, when the phase image is expressed in radians, φ i,j =φ i,j +2π; when the phase image is expressed in degrees, φi,j =φ i,j +360. φ i,j Represents the pixel argument.

[0023] The step 6 specifically includes: calculating pixel by pixel for the phase image φ, or A new data image is formed and recorded as V, that is, when expressed in radians, Or when expressed in degrees,

[0024] The beneficial technical effects of the present invention are:

[0025] 1. Through the polarization synthetic aperture radar complex image data transformation method provided by the present invention, the amplitude and phase of the SAR complex image data are transformed into another two-dimensional vector space, and the new data has almost the same numerical range and statistical distribution form, which is conducive to improving the application degree and depth of phase data in SAR image target enhancement and detection, improving the interpretation accuracy of SAR remote sensing images, and promoting the target classification and detection application of machine learning in the SAR field.

[0026] 2. Through the polarization synthetic aperture radar complex image data transformation method provided by the present invention, the amplitude distribution of the SAR image is distributed between 0 and 1 through transformation. Compared with the logarithmic transformation, no negative values ​​appear, and it has relative stability and accuracy. After the transformation, the phase and pixel values ​​have the same statistical distribution and distribution range, which is convenient for comprehensive consideration. The present invention effectively avoids the following problems existing in the prior art: the amplitude range of general SAR images is relatively large, and logarithmic transformation is often used to represent them. In the logarithmic transformation, the values ​​less than 1 are less than 0 after transformation, thus resulting in negative values, resulting in a large range of numerical changes for the same target in different images, and the specific numerical expressions are negative and positive; the amplitude ranges of different images are also different, which affects the stability of the expression of the same target.

[0027] 3. The polarization synthetic aperture radar complex image data transformation method provided by the present invention increases the difference in the spatial statistical distribution form of the target and its neighborhood pixels, which is beneficial to improving the target distinction and retrieval capabilities.

[0028] 4. The polarization synthetic aperture radar complex image data transformation method provided by the present invention is for transforming single-polarization SAR data, while target polarization transformation analysis usually requires multi-polarization data. Obviously, for target multi-polarization transformation analysis, the present invention provides a more basic data transformation. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1A is the real part image of SAR complex image;

[0030] Figure 1B is the imaginary part of the SAR complex image;

[0031] Figure 2A is the amplitude image;

[0032] Figure 2B is the Rayleigh distribution diagram of the amplitude;

[0033] Figure 3A is the phase image;

[0034] Figure 3B is the uniform distribution diagram of -π to π phase;

[0035] Figure 4A is the U image;

[0036] Figure 4B is the 0-1 uniform distribution map of the U image;

[0037] Figure 5A is the phase distribution range transformation diagram;

[0038] Figure 5B It is a uniform distribution diagram of phase 0~2π;

[0039] Fig. 6A is the phase V image;

[0040] Figure 6B is the uniform distribution 0~1 map of V image;

[0041] Figure 7 This is the SAR complex image data transformation diagram. DETAILED DESCRIPTION

[0042] The present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0043] A method for transforming complex image data of a polarization synthetic aperture radar, the method specifically comprising the following steps:

[0044] In the following image transformation, SNAP is used to read the original RADARSAT-2 data. The specific invention algorithm is implemented using PYTHON+GDAL coding. The data format is recorded in ENVI format. Of course, other formats such as Geotiff can also be used. The image browsing and display are displayed using ENVI software.

[0045] Step 1: Select SAR complex image

[0046] Select the entire SAR complex image or a sub-image containing some pixels, set the number of rows (height or length) to M, the number of columns (width) to N, then the total number of pixels in the image is M×N, and the subscript (i, j) represents the specific position of the pixel in the image, which is the i-th row and the j-th column.

[0047] Step 2: Convert the SAR complex image into amplitude image A and phase image φ

[0048] Euler's formula is used to convert the pixel complex number from algebraic form to complex exponential form to form the amplitude image A and the phase image φ.

[0049] Among them, the amplitude image A corresponds to the modulus of the complex pixel values, and the phase image φ corresponds to the argument of the complex pixel values.

[0050] Step 3: Calculate the mean square error σ of the amplitude image A

[0051] Calculate the mean square error (RMSE) of all pixels in the amplitude image A and record its value as σ. The calculation formula is:

[0052] Where σ is the mean square error of the amplitude image, A i,j is the pixel amplitude value of the i-th row and j-th column, A m is the mean of image A, M is the total number of rows of amplitude image A, and N is the total number of columns of amplitude image A.

[0053] When performing data transformation on a local target and its neighborhood, all calculation parameters are determined based on the local window size parameters M and N. Typical values ​​of M and N are odd numbers such as 3, 5, 7, 9, 11, and 13.

[0054] Step 4: Calculate the U image based on the mean square error σ of the amplitude image A

[0055] For the amplitude image A, calculate pixel by pixel The value of is assigned accordingly to form a new image and recorded as U, that is,

[0056] Step 5: Calculate the phase image φ

[0057] For the phase image φ, calculate pixel by pixel. If φ i,j <0, when the phase image is expressed in radians, φ i,j =φ i,j +2π; when the phase image is expressed in degrees, φ i,j =φ i,j +360. φ i,j Represents the pixel argument.

[0058] Step 6: Calculate the phase V image based on the phase image φ

[0059] For the phase image φ, calculate pixel by pixel, or A new data image is formed and recorded as V, that is, when expressed in radians, Or when expressed in degrees,

[0060] Step 7: The calculation result (U, V) is the final SAR complex image data transformation result.

[0061] According to the U image obtained in step 4 and the phase V image obtained in step 6, the result (U, V) is obtained, which is the final SAR complex image data transformation result.

[0062] In the transformation results, U and V are both non-negative real numbers, and their distribution range is between 0 and 1.

[0063] Example 1

[0064] Taking the data product RADARSAT-2 SLC HH polarization as an example, SAR complex image data transformation is performed

[0065] Step 1: Select SAR complex image

[0066] As shown in Figure 1, a SAR complex image is selected, which consists of a real part and an imaginary part. Figure 1A is the real part image, Figure 1B is the imaginary image, and the complex value of the corresponding pixel in the image is given by Figure 1A and Figure 1B The values ​​of the corresponding points in the image constitute a pair of complex numerical coordinates. In this example, the image height M=6220 and the image width N=3644.

[0067] Step 2: Convert the SAR complex image into amplitude image A and phase image φ

[0068] Euler's formula is used to convert the pixel complex number from algebraic form to complex exponential form to form the amplitude image A and the phase image φ.

[0069] Figure 2 is an amplitude image, where Figure 2A is the amplitude image, Figure 2B is the Rayleigh distribution diagram of the amplitude. Figure 3 is the phase image, where Figure 3A is the phase image, Figure 3B This is a uniform distribution diagram of the phase from -π to π.

[0070] It can be seen from Figure 2-3 that the difference in amplitude distribution and phase distribution and the difference in numerical range are quite large.

[0071] Step 3: Calculate the mean square error σ of the amplitude image A

[0072] Apply the formula Calculate the mean A of the amplitude image m =0.220896, A i,j is the pixel amplitude corresponding to the position i row j column, and then use the formula The mean square error σ of the amplitude image is calculated as 0.158856. In this example, the mean square error of the entire image is calculated.

[0073] Step 4: Calculate the U image based on the mean square error σ of the amplitude image A

[0074] Apply the formula Calculate the U value of the corresponding pixel in the amplitude image. In this example, σ A Represents the mean square error of the entire amplitude image, and its value is obtained by the calculation result in step 3.

[0075] Figure 4 shows the data conversion result U of the amplitude image and its distribution, where Figure 4A is the U image, Figure 4B It is the 0~1 uniform distribution map of U image.

[0076] Step 5: Calculate the phase image φ

[0077] Figure 5 shows the image and its distribution after the phase distribution range is transformed. Figure 5A is the phase distribution range transformation diagram, Figure 5B It is a uniform distribution diagram of phase 0~2π.

[0078] Step 6: Calculate the phase V image based on the phase image φ

[0079] Figure 6 shows the final conversion result V of the phase image and its distribution, where: Fig. 6A is the phase V image, Figure 6B It is the uniform distribution 0~1 map of V image.

[0080] Step 7: The calculation result (U, V) is the final SAR complex image data transformation result.

[0081] The calculation results of step 4 and step 6 constitute the final result. After the data conversion of the method in this paper, U and V have the same value distribution range of 0 to 1 and almost the same statistical distribution form.

[0082] like Figure 7 As shown in Figure 1, a set of SAR complex image data transformation diagrams are provided. Figure 7 In the figure, the statistical features of targets in different areas of the same data image are selected for comparison. The different red frames in the SAR image (middle) represent three targets at different positions. After the target area amplitude data is transformed, it can be seen from the shape of the target histogram that the feature difference of the target after the transformation (left) is significantly greater than that before the transformation (right).

[0083] The present invention is described in detail above with reference to the accompanying drawings and embodiments, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of ordinary technicians in the field without departing from the purpose of the present invention. The contents not described in detail in the present invention can adopt the existing technology.

Claims

1. A method for transforming complex image data of polarization synthetic aperture radar, characterized in that: The method comprises the following steps: Step 1, select SAR complex image; Step 2: Convert the SAR complex image into an amplitude image A and a phase image φ; Step 3, calculate the mean square error σ of the amplitude image A; Step 4: Calculate the U image according to the mean square error σ of the amplitude image A; Step 5, calculate the phase image φ; Step 6, calculating the phase V image according to the phase image φ; Step 7: The calculation result (U, V) is the final SAR complex image data transformation result; The step 4 is specifically as follows: for the amplitude image A, pixel by pixel calculation The value of is assigned accordingly to form a new image and recorded as U, that is, Among them, A i,j is the pixel amplitude value of the i-th row and j-th column; The step 5 is specifically as follows: for the phase image φ, calculate pixel by pixel, if φ i,j <0, when the phase image is expressed in radians, φ i,j =φ i,j +2π; when the phase image is expressed in degrees, φ i,j =φ i,j +360,φ i,j Represents the pixel argument.

2. The method for transforming complex image data of polarization synthetic aperture radar according to claim 1, characterized in that: The step 1 is specifically as follows: select the entire SAR complex image or a sub-image containing some pixels, and assume that the number of rows is M and the number of columns is N, then the total number of pixels in the image is M×N, and the subscript (i, j) represents that the specific position of the pixel in the image is the i-th row and the j-th column.

3. The method for transforming complex image data of polarization synthetic aperture radar according to claim 1, characterized in that: The step 2 specifically includes: using the Euler formula to convert the pixel complex number from an algebraic form to an exponential form of a complex number, thereby forming an amplitude image A and a phase image φ.

4. The method for transforming complex image data of polarization synthetic aperture radar according to claim 1, characterized in that: The formula for calculating the mean square error σ of the amplitude image A in step 3 is: Where σ is the mean square error of the amplitude image, A i,j is the pixel amplitude value of the i-th row and j-th column, A m is the mean of image A, M is the total number of rows of amplitude image A, and N is the total number of columns of amplitude image A.

5. The method for transforming complex image data of polarization synthetic aperture radar according to claim 1, characterized in that: The step 6 specifically includes: calculating pixel by pixel for the phase image φ, A new data image is formed and recorded as V, that is, when expressed in radians, Or when expressed in degrees,

Images