A SAR azimuth ambiguity resolution method based on spectrum asymmetry

Abstract

The application discloses a SAR azimuth ambiguity recognition method based on spectrum asymmetry, comprising the following steps: performing multi-view preprocessing on single-view complex image data of a synthetic aperture radar; performing initial target detection on the single-view complex image data of the synthetic aperture radar after the multi-view preprocessing to obtain an initial detection mask, and performing sub-band division on original single-view complex image data of the synthetic aperture radar; extracting local energy features of the initial detection mask in the divided sub-bands, and calculating a sub-view divergence index; performing real target recognition according to the sub-view divergence index to generate a recognition mask; and performing pixel-level logical AND operation on the initial detection mask and the recognition mask to obtain a pure target detection map. The application has low calculation complexity and strong robustness, can effectively suppress false alarms caused by an unsatisfactory radar antenna pattern and sampling limitations, and is especially suitable for high-precision marine vessel detection and target recognition of a synthetic aperture radar.

Description

Technical Field

[0001] This invention belongs to the field of target detection technology, specifically a SAR azimuth ambiguity identification method based on spectral asymmetry. Background Technology

[0002] Traditional constant false alarm rate (CFAR) detection methods rely primarily on the intensity characteristics of target echoes, making it difficult to effectively distinguish between real vessels and azimuth ambiguities. Due to the radar antenna gain distribution and pulse repetition frequency limitations, azimuth ambiguities are highly similar in intensity to real targets, leading to severe false alarm problems in traditional intensity detectors. Furthermore, existing identification methods based on sub-view analysis often suffer from high computational complexity and insufficient robustness at low signal-to-noise ratios, making it difficult to meet the requirements of high-precision real-time detection. Summary of the Invention

[0003] This invention proposes a SAR azimuth ambiguity identification method based on spectral asymmetry, which effectively suppresses false targets through CFAR initial detection, azimuth sub-band decomposition and sub-view divergence determination.

[0004] The technical solution to achieve the purpose of this invention is: a SAR azimuth ambiguity identification method based on spectral asymmetry, comprising:

[0005] Step 1: Perform multi-view preprocessing on the synthetic aperture radar single-view complex image data;

[0006] Step 2: Perform initial target detection on the preprocessed multi-view synthetic aperture radar single-view complex image to obtain the initial detection mask, and divide the original synthetic aperture radar single-view complex image data into subbands;

[0007] Step 3: Extract the local energy features of the initial detection mask in the divided sub-bands and calculate the sub-view divergence index; perform real target recognition based on the sub-view divergence index to generate the identification mask;

[0008] Step 4: Perform a pixel-level logical AND operation between the initial detection mask and the identification mask to obtain a clean target detection map.

[0009] Compared with the prior art, the significant advantages of this invention are:

[0010] 1. This invention extracts the target's azimuth Doppler spectral features and utilizes sub-view divergence. This parameter quantitatively characterizes spectral asymmetry. Compared to traditional methods that rely solely on intensity, it can effectively identify and eliminate [specific asymmetries]. The value tends to 1, indicating the azimuth towards the false target, thus preserving the real ship ( While approaching 0, it significantly reduces systemic false alarms.

[0011] 2. This invention establishes an identification threshold. Equivalent number of views of an image and expected false negative rate The relationship between them is explained. By introducing a regularized incomplete Beta function to adaptively determine the threshold, the threshold setting is theoretically grounded and exhibits better robustness.

[0012] 3. The sub-band domain processing framework proposed in this invention operates directly in the frequency domain, with key performance indicators... The calculations involve only addition, subtraction, multiplication, and division of local energies, resulting in low computational complexity and good real-time performance. Compared to existing complex statistical modeling and identification methods, the computational overhead is relatively small.

[0013] 4. This invention employs a pixel-level logical fusion mechanism to integrate the identification mask. Compared with the initial detection image Perform an AND operation. This process can be integrated as a standalone plugin into various constant false alarm rate (CFAR) detection algorithms to achieve refined performance improvements without altering the original system architecture.

[0014] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0015] Figure 1 This is a diagram illustrating the effect of processing synthetic aperture radar image data according to the present invention.

[0016] Figure 2 This is a flowchart of the present invention. Detailed Implementation

[0017] like Figure 2 As shown, a SAR azimuth ambiguity identification method based on spectral asymmetry includes:

[0018] Step 1: Perform multi-look preprocessing on the synthetic aperture radar (SAR) single-look complex (SLC) image data, specifically as follows:

[0019] Intensity data of SLC images Intensity multi-view averaging is used to reduce speckle noise and decrease the amount of computational data. Among these, For directional multi-view numbers, For range-oriented multi-view, select an appropriate [method / approach] based on the row and column resolution ratio of the SLC image. and The value is set to make the pixels of the multi-view image tend to be square, or according to the processing efficiency requirements. and Equal (e.g.) or ).

[0020] Step 2: Perform initial target detection on the preprocessed multi-view synthetic aperture radar (SAR) single-view complex image, and divide the SAR single-view complex (SLC) image data into subbands, specifically as follows:

[0021] Initial target detection: Using a cell-average constant false alarm rate (CA-CFAR) detector, based on a preset false alarm rate (e.g. Given the background window size, obtain an initial detection mask containing the real ship and the orientation-blurred pixels. ;

[0022] Azimuth sub-band division: The original single-view complex data (SLC) is transformed into the Doppler frequency domain using a azimuth-oriented Fast Fourier Transform (FFT) to obtain the Doppler spectrum. With the Doppler center frequency Using this as a boundary, the Doppler spectrum is divided. It is divided into two symmetrical sub-bands, upper and lower.

[0023]

[0024]

[0025] In the formula, This represents the azimuth-direction Doppler frequency variable.

[0026] Step 3: Extract the initial detection mask The local energy characteristics in the two sub-views are analyzed, and the sub-view divergence index, which quantitatively characterizes spectral asymmetry, is dynamically calculated using the ratio of the absolute value of the energy difference to the sum of the energies. Specifically:

[0027] For the upper and lower sub-bands and The first subview is generated by performing inverse Fourier transforms on each of the following methods. Second subview , , These are the azimuth and range indices, respectively.

[0028] For the initial detection mask For suspected pixels in the data, a spatial sliding window (e.g.) is used. Local energy estimation is performed within a window, and the sub-view divergence index is calculated. Sub-view divergence Defined as the ratio of the absolute value of the local energy difference between two subviews to the total energy, specifically:

[0029]

[0030] In practical implementation, by using pixels A two-dimensional sliding window is set around the center, and the mean value of the squared magnitude of all pixels within the window is calculated to suppress speckle noise and obtain robust local energy estimation.

[0031] This index utilizes the asymmetric physical properties of orientation ambiguity in the Doppler spectrum: the energy of the real target is... and The distribution is relatively balanced. The energy approaches 0; however, the energy of azimuth-blurred images is typically concentrated in a single sub-band, leading to... Approaching 1.

[0032] Using the calculated Perform fuzzy judgment: based on a preset false negative rate (e.g.) The recognition threshold is determined by the number of image views. Retain indicators Pixels below the recognition threshold are treated as real targets, and a recognition mask is generated. ,Right now:

[0033] like ,but (Identified as a real ship)

[0034] like ,but (Judged as a vague false alarm)

[0035] Specifically, in order to scientifically set thresholds to control the false negative rate Based on the probability density function, the following is derived: The calculation formula is as follows:

[0036]

[0037] in, For a regularized incomplete Beta function, its upper limit variable for integration is... , Let be the equivalent number of views of the image. Using this formula, given the number of views, the optimal identification threshold can be adaptively calculated based on the allowable false negative rate. The calculation formula is as follows:

[0038]

[0039] in This is the inverse incomplete Beta function, defined by setting the allowed false negative rate. and equivalent visual number Thus, the corresponding threshold can be calculated. .

[0040] Step 4: Result Fusion and Output. The initial detection mask... With identification mask Perform pixel-level logical AND operations, i.e. To eliminate The system generates fuzzy false alarm points from the azimuth to the side lobes; finally, it outputs the processed, clean ship detection map.

[0041] Figure 1 This is a diagram illustrating the effect of processing synthetic aperture radar image data according to the present invention. Green represents real ships, and red represents azimuth-ambiguous false targets. Figure 1 The detection results show that in complex SAR sea surface observation scenarios with significant strong scattering interference, the identification method based on spectral asymmetry proposed in this invention can accurately identify azimuthally ambiguous false targets (red boxes) that are extremely similar in intensity to real targets, and effectively remove them from the initial CFAR detection results. Simultaneously, the algorithm completely and accurately preserves the real vessel targets (green boxes). This intuitively demonstrates that this invention possesses extremely strong systematic false alarm suppression capabilities without sacrificing the false target miss rate, significantly improving the reliability of high-precision maritime target detection.

Claims

1. A SAR azimuth ambiguity identification method based on spectral asymmetry, characterized in that, include: Step 1: Perform multi-view preprocessing on the synthetic aperture radar single-view complex image data; Step 2: Perform initial target detection on the preprocessed multi-view synthetic aperture radar single-view complex image to obtain the initial detection mask, and divide the original synthetic aperture radar single-view complex image data into subbands; Step 3: Extract the local energy features of the initial detection mask in the divided sub-bands and calculate the sub-view divergence index. The specific method is as follows: Perform inverse Fourier transforms on the upper and lower sub-bands respectively to generate the first sub-view. Second subview , , These are the azimuth and range indices, respectively. Using a spatial sliding window for the initial detection mask Local energy estimation is performed on suspected pixels in the image, and the sub-view divergence index is calculated. Specifically: ； In the formula, Represents a spatial local averaging operator; Based on the sub-view divergence index, real target identification is performed, and a recognition mask is generated. Step 4: Perform a pixel-level logical AND operation between the initial detection mask and the identification mask to obtain a clean target detection map.

2. The SAR azimuth ambiguity identification method based on spectral asymmetry according to claim 1, characterized in that, The initial target detection for the preprocessed synthetic aperture radar single-view complex image is as follows: using a cell-average constant false alarm rate detector, based on a preset false alarm rate and background window size, an initial detection mask containing real ships and azimuth-ambiguous pixels is obtained. .

3. The SAR azimuth ambiguity identification method based on spectral asymmetry according to claim 1, characterized in that, The specific method for subbanding the raw synthetic aperture radar single-view complex image data is as follows: Azimuth-oriented fast Fourier transform is performed on the original synthetic aperture radar single-look complex image data to obtain the Doppler spectrum; The Doppler spectrum is divided into two symmetrical sub-bands, upper and lower, with the Doppler center frequency as the boundary.

4. The SAR azimuth ambiguity identification method based on spectral asymmetry according to claim 1, characterized in that, The specific method for generating a recognition mask based on the sub-view divergence index to identify the real target is as follows: The identification threshold is determined based on the preset false negative rate and the number of image views. Pixels with subview divergence indices below the recognition threshold are retained as real targets to generate a recognition mask. .

5. The SAR azimuth ambiguity identification method based on spectral asymmetry according to claim 4, characterized in that, The identification threshold is determined based on the preset false negative rate and the number of image views. The specific formula is: ； In the formula, This is the inverse incomplete Beta function. To preset the false negative rate, This is the equivalent apparent number.

Images