Small fishing boat detection method using dual polarization channel data

Abstract

The present invention relates to a method for detecting small fishing vessels using dual-polarization channel data. The small fishing vessel detection method of the present invention receives Synthetic Aperture Radar (SAR) Sentinel-1 Single Look Complex (SLC) data as input, preprocesses the SAR SLC data to obtain VH polarization channel data (SVH) and VV polarization channel data (SVV), fuses the VH polarization channel data (SVH) and VV polarization channel data (SVV) to generate polarization fusion image data (fusVH), generates cross-correlation image data (sym) based on the VH polarization channel data (SVH) and VV polarization channel data (SVV), fuses the polarization fusion image data (fusVH) and cross-correlation image data (sym) to generate fusion image data (fussym), classifies vessels into commercial vessels and small fishing vessels based on RCS (backscatter intensity), filters out commercial vessels to determine small fishing vessel detection candidates, and the polarization fusion image data (fusVH) and fusion image It is configured to adaptively apply a threshold based on the target type for each data (fussym), and to determine small fishing vessels by filtering commercial vessels from the small fishing vessel detection candidates using the polarized fused image data (fusVH) and fused image data (fussym) to which the adaptive threshold has been applied.

Description

Technology Field

[0001] The present invention relates to a method for detecting small fishing vessels using dual-polarization channel data, and more specifically, to a method for detecting small fishing vessels using VH polarization (cross-polarization) channel data and VV polarization (co-polarization) channel data obtained by preprocessing SAR (Synthetic Aperture Radar) SLC (Sentinel-1 Single Look Complex) data [e.g., Sentinel-1 SLC data acquired from a Sentinel-1 satellite]. Background Technology

[0002] Maritime Domain Awareness (MDA) is essential for ensuring maritime security against threats such as illegal vessels, illegal fishing activities, and maritime accidents. Various data can be utilized to enhance MDA, and among them, Synthetic Aperture Radar (SAR) systems are widely used because they can observe a wide range of ocean areas and provide ocean surface data unaffected by weather conditions.

[0003] Radar pulses emitted from SAR satellites mostly interact with the ocean surface through single-bounce scattering, but objects such as ships typically generate double or multiple bounces due to their complex superstructures and reflective surfaces. Consequently, the backscattering intensity of a ship appears much stronger than that of the surrounding environment, which is advantageous for detecting ships through various approaches.

[0004] However, various factors such as ship characteristics, environmental conditions, radar characteristics, image quality, and spatial resolution affect the detection effectiveness of SAR-based ships. Furthermore, a comparative analysis of ship detection effectiveness in RADARSAT-1 beam mode and ERS-1 / 2 imagery revealed that detection performance improves as wind speed decreases, the angle of incidence increases, and image resolution increases. However, detecting small vessels, which are smaller than the spatial resolution of satellite imagery, remains a difficult challenge.

[0005] Over the past few decades, various ship detection techniques have been proposed. Among them, the Constant False Alarm Rate (CFAR) technique has been widely used because ships generally exhibit relatively stronger backscattering than background ocean clutter. In particular, Cell Averaging CFAR (CA-CFAR) is a simple technique that assumes the clutter distribution is Gaussian and performs detection using Boxcar filtering.

[0006] Since then, these methods have evolved with the emergence of multi-satellite platforms providing various resolutions. Additionally, image enhancement techniques combining feature extraction and CFAR methods have been proposed to achieve high ship detection performance. Nevertheless, small vessels remain a very difficult challenge to detect due to their low backscattering intensity.

[0007] In addition to traditional techniques, recent advancements in machine learning, particularly Convolutional Neural Networks (CNNs), have presented new possibilities for ship detection. Researchers have successfully performed ship detection experimentally by applying advanced CNN-based models, such as Region-Based CNN (R-CNN), Single Shot MultiBox Detector (SSDD), and You Only Look Once (YOLO), to satellite imagery.

[0008] Large datasets, such as ship chip images collected through long-term satellite observations, have been built to support these models, and open-source deep learning models like xView3-SAR have greatly facilitated the training and inference processes for ship detection.

[0009] However, CNN models still face limitations in detecting small vessels in SAR imagery. Since small vessels have a very low pixel count in SAR images and lack distinct image features, there is a problem where the model's detection performance is significantly degraded.

[0010] Meanwhile, SAR-based ship detection faces problems with radio frequency interference (RFI) and azimuth ambiguities / smearing. Ambiguity is caused by strong backscattering from artificial structures, and RFI is triggered by human activities such as mobile communication systems or other satellites, which can lead to false detections.

[0011] Azimuth smearing occurs due to the movement of the detection target, resulting in a discrepancy between the target's expected Doppler frequency and the actual observed Doppler frequency. Consequently, the target's energy spreads over a wide area, reducing the peak energy and negatively impacting detection accuracy.

[0012] Accordingly, a method was proposed to suppress RFI and azimuth blurring by combining co-polarized and cross-polarized images from dual-polarized Sentinel-1 GRD data

[20] . However, since this method has been studied mainly on ships equipped with Automatic Identification Systems (AIS), it has limitations in detecting small vessels.

[0013] Various targets generally interact with plane waves of different polarizations, which provide characteristic information regarding amplitude and phase in polarized SAR (PolSAR). Based on these characteristics, various ship detection techniques have been developed.

[0014] The polarimetric reflection symmetry detector is based on the principle that natural environments possess reflection symmetry. On the other hand, complex scatterers such as ships do not possess reflection symmetry. This detector is applied to the cross-correlation between co-polarized and cross-polarized channels.

[0015] The Polarimetric Notch Filter (PNF) technique focuses on the second-order polarization characteristic vector of a ship located in an orthogonal complement space with respect to the sea.

[0016] The Polarimetric Match Filter (PMF) is a method that enhances the contrast of the covariance matrix between ocean clutter and ships by considering different scattering mechanisms. Similarly, the Polarimetric Whitening Filter (PWF) and the Optimal Polarimetric Detector (OPD) have been developed to minimize speckle and are applicable in marine environments.

[0017] Sub-look detectors focus on evaluating correlations between portions of the image spectrum in single-look SAR data. This is because ships are expected to be cross-correlated, while the ocean is not.

[0018] The Polarimetric Entropy Detector considers the randomness of the scattering mechanism forming the target and assumes that the ocean is relatively polarized because it is always the surface. This detector is particularly effective at low angles of incidence.

[0019] The Dual-pol Ratio Anomaly Detector (DpolRAD) supports asynchronous dual-polation SAR data such as Sentinel-1 and is designed for iceberg detection by monitoring anomalies in volume scattering surrounded by surface scattering.

[0020] Most existing polarization SAR research has centered on satellite platforms providing quad-polarization data. In contrast, research on polarization SAR applications for Sentinel-1, which operates in dual-polarization, has not yet been sufficiently developed. Furthermore, most existing studies focus on commercial vessels equipped with AIS, and qualitative research on the detection of small vessels is relatively lacking.

[0021] To bridge this research gap, recent studies have been conducted on the detection of small vessels (recreational boats) using Sentinel-1 data by applying mathematical morphological operators and threshold-based methods. Additionally, experimental detection studies targeting non-metallic rubber boats have been carried out using dual-polarized Sentinel-1 data.

[0022] However, most of these studies have been conducted in inland waters, so there are limitations to their application in open sea environments. Prior art literature

[0023] Republic of Korea Published Patent Application No. 10-2004-0075488 (Title of Invention: Automatic Identification and Monitoring System for Fishing Vessels) The problem to be solved

[0024] Accordingly, the present invention has been made to solve the problems of the above situation, and the objective of the present invention is to provide a method for detecting small fishing vessels using dual-polarized channel data that can stably detect small fishing vessels with low backscattering intensity. means of solving the problem

[0025] To achieve the above objective, a method for detecting small fishing vessels using dual-polarization channel data according to an embodiment of the present invention comprises the step of inputting SAR (Synthetic Aperture Radar) SLC (Sentinel-1 Single Look Complex) data by a SAR (Synthetic Aperture Radar) SLC (Sentinel-1 Single Look Complex) data input unit; and a preprocessing unit preprocessing the SAR SLC data to obtain VH polarization (cross-polarization) channel data (S VH ) and VV polarization (co-polarization) channel data (S VV A step of acquiring ); a polarization channel fusion unit acquiring the VH polarization channel data (S VH ) and VV polarized channel data(S VV A step of generating polarized fused image data (fusVH) by fusing ); a cross-correlated image data generation unit generating the VH polarized channel data (S VH ) and VV polarized channel data(S VV The method is characterized by comprising: a step of generating cross-correlated image data (sym) based on ); a step in which a fusion image data generation unit fuses the polarized fusion image data (fusVH) and the cross-correlated image data (sym) to generate fusion image data (fussym); a step in which a vessel separation unit classifies commercial vessels and small fishing vessels based on RCS (backscatter intensity) and filters commercial vessels to determine small fishing vessel detection candidates; a step in which an adaptive threshold application unit adaptively applies a threshold based on the target type to each of the polarized fusion image data (fusVH) and the fusion image data (fussym); and a step in which a small fishing vessel determination unit determines small fishing vessels by filtering commercial vessels from the small fishing vessel detection candidates using the polarized fusion image data (fusVH) and the fusion image data (fussym) to which the threshold has been adaptively applied.

[0026] The small fishing vessel detection method using dual-polarized channel data according to the above embodiment may further include a step in which an average filtering unit adds average filtered image data of the real and imaginary components of the fused image data (fussym) to the fused image data (fussym) to suppress sea clutter.

[0027] In the method for detecting small fishing vessels using dual-polarization channel data according to the above embodiment, the polarization fusion image data (fusVH) generation step is the VH polarization channel data (S VH ) and the above VV polarization channel data (S VV VH polarization channel strength (|S) from ) VH |²) and VV polarization channel strength (|S VV The method may include a step of extracting |²); and a step in which, if the difference between the VV polarized channel strength and the VH polarized channel strength is less than 6.53 dB, a value obtained by subtracting 6.53 dB from the VV polarized channel strength is used, while if the difference between the VV polarized channel strength and the VH polarized channel strength is 6.53 dB or more, the VH polarized channel strength is used.

[0028] In a method for detecting small fishing vessels using dual polarization channel data according to the above embodiment, the step of generating the fusion image data (fussym) may include: a step of extracting polarization fusion image intensity and cross-correlation image intensity from the polarization fusion image data (fusVH) and cross-correlation image data (sym); and a step of using the larger value (Max) between the cross-correlation image intensity and the polarization fusion image intensity when the difference between the cross-correlation image intensity and the polarization fusion image intensity is less than 7.2 dB and the cross-correlation image intensity is greater than -18.03 dB, while using the polarization fusion image intensity when the difference between the cross-correlation image intensity and the polarization fusion image intensity is 7.2 dB or more and the cross-correlation image intensity is -18.03 dB or less.

[0029] In the method for detecting small fishing vessels using dual-polarized channel data according to the above embodiment, the step of determining a candidate for small fishing vessel detection may include a step of classifying a target as a commercial vessel when the target's RCS is -9.49 dB or higher, and determining a target as a small fishing vessel when the target's RCS is less than -9.49 dB. Effects of the invention

[0030] According to the small fishing vessel detection method using dual-polarized channel data according to an embodiment of the present invention, by using dual-polarized channel data to effectively detect small fishing vessels with low backscattering intensity, it suppresses the effects of fog, interference, marine clutter, and commercial vessels with high backscattering intensity, and has an excellent effect of greatly improving detection reliability through an automated adaptive detection process. Brief explanation of the drawing

[0031] FIG. 1 is a block diagram of a small fishing boat detection system using dual-polarized channel data according to an embodiment of the present invention. Figure 2 is a diagram showing the total number of ships observed by V-PASS when collecting data in the ROIS (Region of Interest) along with the list of Sentinel-1 SLC (Sentinel-1 Single Look Complex) data used in the present invention. FIG. 3 is a flowchart for explaining a method for detecting small fishing boats using dual polarization channel data according to an embodiment of the present invention. FIG. 4 is a diagram illustrating the process of generating fussym data for detecting small fishing boats according to an embodiment of the present invention. FIG. 5 is a diagram illustrating the commercial vessel detection results for (a) ROI1 and (b) ROI5 according to an embodiment of the present invention. Specific details for implementing the invention

[0032] In describing the embodiments of the present invention, if it is determined that a detailed description of known technology related to the present invention may unnecessarily obscure the essence of the present invention, such detailed description will be omitted. Furthermore, the terms described below are defined considering their functions in the present invention, and these may vary depending on the intentions or practices of the user or operator. Therefore, such definitions should be based on the content throughout this specification. Terms used in the detailed description are intended merely to describe the embodiments of the present invention and should not be interpreted restrictively. Unless explicitly stated otherwise, expressions in the singular form include the meaning of the plural form. In this description, expressions such as "include" or "comprise" are intended to refer to certain characteristics, numbers, steps, actions, elements, parts thereof, or combinations thereof, and should not be interpreted as excluding the existence or possibility of one or more other characteristics, numbers, steps, actions, elements, parts thereof, or combinations thereof other than those described.

[0033] In each system illustrated in the drawings, elements in some cases may have the same or different reference numbers, suggesting that the represented elements may be different or similar. However, elements may have different implementations and may operate with some or all of the systems shown or described herein. The various elements illustrated in the drawings may be the same or different. It is optional which is referred to as the first element and which is referred to as the second element.

[0034] In this specification, the phrase “transmits,” “delives,” or “provides” data or signals from one component to another component includes not only the direct transmission of data or signals from one component to another component, but also the transmission of data or signals to another component through at least one other component.

[0035] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[0036] FIG. 1 is a block diagram of a small fishing boat detection system using dual-polarized channel data according to an embodiment of the present invention.

[0037] A small fishing vessel detection system using dual polarization channel data according to an embodiment of the present invention includes, as illustrated in FIG. 1, a SAR SLC data input unit (100), a preprocessing unit (200), a polarization channel fusion unit (300), a cross-correlation image data generation unit (400), a fusion image data generation unit (500), an average filtering unit (600), a vessel separation unit (700), an adaptive threshold application unit (800), and a small fishing vessel determination unit (900). These components may be integrated on a single terminal device (e.g., a laptop, personal computer, PDA, PMP, smartphone, etc.) or processor, or configured as separate devices.

[0038] The SAR SLC data input unit (100) serves to input SAR SLC data (e.g., Sentinel-1 SLC data). Sentinel-1 SLC data refers to Synthetic Aperture Radar (SAR) observation data acquired from the Sentinel-1 satellite of the Copernicus program operated by the European Space Agency (ESA) and the European Union (EC), and is radar satellite image data that provides high-resolution Earth surface information regardless of weather conditions. FIG. 2 illustrates a list of Sentinel-1 SLC (Sentinel-1 Single Look Complex) data used in the present invention, along with the total number of vessels observed by V-PASS (Small Fishing Vessel Position Tracking System) when collecting data in the ROIS (Region of Interest).

[0039] The preprocessing unit (200) preprocesses the SAR SLC data input by the SAR SLC data input unit (100) to obtain VH cross-polarization channel data (S VH )(also called VH polarized complex signal) and VV polarized (co-polarization) channel data (S VV It serves the role of acquiring the )(also called the VV polarized complex signal). Preprocessing can be performed using the **Sentinel Application Platform (SNAP, version 9.0) distributed by the European Space Agency (ESA)**, and the preprocessing process may include TOPSAR segmentation, radiation correction, deburst, scattering matrix calculation, multilook setting, and ellipsoid correction.

[0040] The polarization channel fusion unit (300) is VH polarization channel data (S VH ) and VV polarized channel data(S VV It plays the role of generating polarized fused image data (fusVH) by fusing )

[0041] The process of generating polarized fused image data (fusVH) is as follows.

[0042] First, VH polarization channel data (S VH ) and VV polarized channel data(S VV VH polarization channel strength (|S) from ) VH |²) and VV polarization channel strength (|S VV Extract |²)

[0043] Next, VV polarization channel strength (|S VV |²) and VH polarization channel strength (|S VH If the difference in |²) is less than 6.53 dB, the VV polarization channel strength (|S VV While the value obtained by subtracting 6.53 dB from |²) is used (see Equation 1), the VV polarization channel strength (|S VV |²) and VH polarization channel strength (|S VH If the difference in |²) is 6.53 dB or greater, the VH polarization channel strength (|S VH |²) is used.

[0044] [Mathematical Formula 1]

[0045]

[0046] The cross-correlation image data generation unit (400) generates VH polarization channel data (S VH ) and VV polarized channel data(S VV It plays the role of generating cross-correlated image data (sym) based on ).

[0047] Cross-correlated image data (sym) can be calculated by the following [Equation 2].

[0048] [Mathematical Formula 2]

[0049]

[0050] [Here, *: complex conjugate, <·>: spatial or local mean (3x3 window, etc.)]

[0051] The fused image data generation unit (500) plays the role of generating fused image data (fussym) by fusing polarized fused image data (fusVH) and cross-correlated image data (sym).

[0052] This explains the process of generating fused image data (fussym).

[0053] First, polarization fusion image intensity and cross-correlation image intensity are extracted from polarization fusion image data (fusVH) and cross-correlation image data (sym).

[0054] Next, if the difference between the extracted cross-correlation image intensity and the polarization fusion image intensity is less than 7.2 dB and the cross-correlation image intensity is greater than -18.03 dB, the larger value (Max) between the cross-correlation image intensity and the polarization fusion image intensity is used (see [Equation 3]).

[0055] Meanwhile, if the difference between the cross-correlation image intensity and the polarization fusion image intensity is 7.2 dB or more and the cross-correlation image intensity is -18.03 dB or less, the polarization fusion image intensity is used.

[0056] [Mathematical Formula 3]

[0057]

[0058] The average filtering unit (600) plays a role in suppressing sea clutter by adding the average filtered image data of the real and imaginary components of the fused image data (fussym) generated by the fused image data generation unit (500) to the fused image data (fussym).

[0059] The vessel separation unit (700) distinguishes between commercial vessels and small fishing vessels based on the RCS (Radar Cross Section) (backscattering intensity), and filters out commercial vessels to determine candidates for detecting small fishing vessels. For example, if the RCS of a target is -9.49 dB or higher, it is classified as a commercial vessel, and if the RCS of a target is less than -9.49 dB, it is judged as a small fishing vessel and can be determined as a candidate small fishing vessel.

[0060] The adaptive threshold application unit (800) is responsible for adaptively applying a threshold based on the target type (commercial vessel, small fishing vessel) for each of the polarized fused image data (fusVH) and fused image data (fussym).

[0061] For example,

[0062] fusVH (Merchant Vessel Detection): PFA (Probability of False Alarm) = 26

[0063] fussym(Small fishing boat detection): PFA = 48

[0064] (Here, as PFA increases, the threshold decreases and sensitivity increases)

[0065] The small fishing vessel determination unit (900) determines small fishing vessels by filtering commercial vessels from small fishing vessel detection candidates determined by the vessel separation unit (700) using polarized fusion image data (fusVH) and fusion image data (fussym) to which a threshold is applied adaptively. The small fishing vessel determination unit (900) can reduce the false positive rate and increase detection precision by comparing and analyzing the adaptive detection results of the polarized fusion image data (fusVH) and fusion image data (fusSym) for targets classified as small fishing vessel candidates by the vessel separation unit (700), excluding targets detected in fusVH by considering them as merchant vessels, and finally determining targets detected only in fusSym as small fishing vessels.

[0066] A method for detecting small fishing boats using dual-polarization channel data, which can be implemented by a small fishing boat detection system using dual-polarization channel data according to an embodiment of the present invention configured as above, will be described with reference to the drawings.

[0067] FIG. 3 is a flowchart for explaining a method for detecting small fishing boats using dual polarization channel data according to an embodiment of the present invention.

[0068] First, the SAR SLC data input unit (100) inputs SAR SLC data [e.g., Sentinel-1 SLC data (radar satellite image data) received from the Sentinel-1 satellite] (S10).

[0069] Next, the preprocessing unit (200) preprocesses the SAR SLC data input by step (S10) to obtain VH cross-polarization channel data (S VH )(also called VH polarized complex signal) and VV polarized (co-polarization) channel data (S VV )(also called VV polarized complex signal) is acquired (S20). The preprocessing steps include a TOPSAR (Terrain Observation by Progressive Scans) segmentation step, a radiation correction step, a deburst step, a scattering matrix calculation step, a multilook setting step, and an ellipsoid correction step, which are described in detail as follows.

[0070] ● TOPSAR (Terrain Observation by Progressive Scans) Segmentation Step

[0071] Sentinel-1 (a satellite platform that captures and generates SAR data) captures in TOPSAR mode and acquires SAR SLC data by dividing it into multiple sub-swaths (partial observation intervals). In this stage, only specific sub-swaths necessary for analysis are extracted to improve processing efficiency.

[0072] ● Radiation correction step

[0073] Since SAR SLC data consists of raw radar intensity values, they must be converted into physical units (e.g., sigma nought, dB). This step standardizes the data to enable comparison between different observation points. Consequently, it is transformed into pixel values ​​that have physical meaning based on the reflective characteristics of the actual surface.

[0074] ● Deburst phase

[0075] TOPSAR collects data in burst mode, consisting of multiple short data bursts. Diburst is the process of reconstructing these bursts into a single continuous image. It aligns the chronological order and restores the spatial connectivity of the images.

[0076] ● Scattering matrix calculation step

[0077] This is a foundational step for extracting scattering characteristics from complex data of different polarizations, such as VH and VV. The scattering matrix or scattering coefficients can be calculated using the polarization response of signals received by SAR. This matrix provides key information for analyzing surface characteristics, such as topography, structures, and vegetation.

[0078] ● Multi-look setup steps

[0079] SAR SLC data exhibits strong speckle noise. To reduce this, the multi-look technique is used. By averaging multiple observation lines, noise is reduced and visual quality is improved at the expense of slightly sacrificing spatial resolution. Generally, this is processed by setting the look number (e.g., range 5 looks, azimuth 1 look, etc.).

[0080] ● Ellipsoid Correction Step

[0081] SAR SLC data is a slant distance-based image acquired from a satellite's viewpoint. The process of converting this into an interpretable image by projecting it onto the geographic coordinate system (latitude / longitude) of the Earth's surface is called ellipsoidal correction. Digital terrain models (DTM or DEM) are also used to correct positional distortion.

[0082] Next, the polarization channel fusion unit (300) VH polarization channel data (S VH ) and VV polarized channel data(S VV ) is fused to generate polarized fused image data (fusVH) (S30). The step (S30) of generating polarized fused image data (fusVH) is as follows.

[0083] First, VH polarization channel data (S VH ) and VV polarized channel data(S VV VH polarization channel strength (|S) from ) VH |²) and VV polarization channel strength (|S VV Extract |²). Next, extract the VV polarization channel strength (|S VV |²) and VH polarization channel strength (|S VH If the difference in |²) is less than 6.53 dB, the VV polarization channel strength (|S VV While the value obtained by subtracting 6.53 dB from |²) is used (see Equation 1), the VV polarization channel strength (|S VV |²) and VH polarization channel strength (|S VH If the difference in |²) is 6.53 dB or greater, the VH polarization channel strength (|S VH |²) is used.

[0084] [Mathematical Formula 1]

[0085]

[0086] Next, the cross-correlation image data generation unit (400) generates VH polarization channel data (S VH ) and VV polarized channel data(S VV Based on ), cross-correlated image data (sym) is generated (S40).

[0087] Cross-correlated image data (sym) can be calculated by the following [Equation 2].

[0088] [Mathematical Formula 2]

[0089]

[0090] [Here, *: complex conjugate, <·>: spatial or local mean (3x3 window, etc.)]

[0091] Next, the fused image data generation unit (500) fuses the polarized fused image data (fusVH) generated in step (S30) and the cross-correlation image data (sym) obtained in step (S40) to generate fused image data (fussym) (S50). The step of generating fused image data (fussym) is described. First, polarized fused image intensity and cross-correlation image intensity are extracted from the polarized fused image data (fusVH) and the cross-correlation image data (sym). Next, if the difference between the extracted cross-correlation image intensity and the polarized fused image intensity is less than 7.2 dB and the cross-correlation image intensity is greater than -18.03 dB, the larger value (Max) between the cross-correlation image intensity and the polarized fused image intensity is used (see [Equation 3]). Meanwhile, if the difference between the cross-correlation image intensity and the polarized fused image intensity is 7.2 dB or more and the cross-correlation image intensity is -18.03 dB or less, the polarized fused image intensity is used.

[0092] [Mathematical Formula 3]

[0093]

[0094] FIG. 4 is a diagram illustrating the process of generating fussym data for detecting small fishing vessels according to an embodiment of the present invention, wherein images (a-1) through (a-5) represent VH polarization channel data, VV polarization channel data, fusVH data, sym data, and final fussym data, respectively. Here, it can be seen that the RCS improves as one moves from (a-1) to (a-5).

[0095] Next, the average filtering unit (600) adds the average filtered image data of the real and imaginary components of the fused image data (fussym) generated by step (S50) to the fused image data (fussym) to suppress sea clutter (S60).

[0096] Next, the ship separation unit (700) separates commercial vessels and small fishing vessels based on the RCS (Radar Cross Section) (backscattering intensity), and filters out commercial vessels to determine candidates for detecting small fishing vessels (S70).

[0097] Next, the adaptive threshold application unit (800) adaptively applies a threshold based on the target type (commercial vessel, small fishing vessel) to each of the polarized fusion image data (fusVH) and fusion image data (fussym) (S80).

[0098] Next, the small fishing vessel determination unit (900) determines the small fishing vessel by filtering commercial vessels from the small fishing vessel detection candidates determined by the vessel separation unit (700) using the polarized fusion image data (fusVH) and fusion image data (fussym) to which a threshold value is adaptively applied by step (S80) (S90).

[0099] FIG. 5 illustrates the results of commercial vessel detection for (a) ROI1 and (b) ROI5 according to an embodiment of the present invention. Here, in ROI1(a), four small fishing vessels and one merchant vessel were observed, and the merchant vessel was a tanker with a length of 37m and a width of 8m. In ROI2(b), two small fishing vessels and one merchant vessel were found, and the merchant vessel was a Ro-Ro / container ship with a length of 143m and a width of 23m. This demonstrates that the merchant vessel was successfully detected.

[0100] According to the small fishing vessel detection method using dual-polarized channel data according to an embodiment of the present invention, by using dual-polarized channel data to effectively detect small fishing vessels with low backscattering intensity, the influence of fog, interference, marine clutter, and commercial vessels with high backscattering intensity is suppressed, and detection reliability can be greatly improved through an automated adaptive detection process.

[0101] Optimal embodiments have been disclosed in the drawings and specification, and specific terms have been used, but these are used only for the purpose of describing embodiments of the invention and are not intended to limit the meaning or the scope of the invention as described in the claims. Therefore, those skilled in the art will understand that various modifications and equivalent alternative embodiments are possible therefrom. Accordingly, the true technical scope of protection of the invention should be determined by the technical spirit of the appended claims. Explanation of the symbols

[0102] 100: SAR SLC Data Input Section 200: Preprocessing section 300: Polarized channel fusion section 400: Cross-correlation image data generation unit 500: Fusion Image Data Generation Unit 600: Average filtering section 700: Ship separation section 800: Adaptive threshold application section 900: Small fishing vessel decision unit

Claims

Claim 1 A method for detecting small fishing vessels using dual-polarization channel data, comprising: a step in which SAR (Synthetic Aperture Radar) SLC (Sentinel-1 Single Look Complex) data is input by a SAR (Synthetic Aperture Radar) SLC (Sentinel-1 Single Look Complex) data input unit (100); and a preprocessing unit (200) preprocessing the SAR SLC data to obtain VH polarization (cross-polarization) channel data (S VH ) and VV polarization (co-polarization) channel data (S VV A step of acquiring ); a polarization channel fusion unit (300) acquiring the VH polarization channel data (S VH ) and VV polarized channel data(S VV A step of generating polarized fused image data (fusVH) by fusing ); a cross-correlated image data generation unit (400) generating the VH polarized channel data (S VH ) and VV polarized channel data(S VV A method for detecting small fishing vessels, comprising: a step of generating cross-correlated image data (sym) based on ); a step in which a fusion image data generation unit (500) fuses the polarized fusion image data (fusVH) and the cross-correlated image data (sym) to generate fusion image data (fussym); a step in which a vessel separation unit (700) classifies commercial vessels and small fishing vessels based on RCS (backscatter intensity) and filters commercial vessels to determine small fishing vessel detection candidates; a step in which an adaptive threshold application unit (800) adaptively applies a threshold based on a target type to each of the polarized fusion image data (fusVH) and the fusion image data (fussym); and a step in which a small fishing vessel determination unit (900) determines small fishing vessels by filtering commercial vessels from the small fishing vessel detection candidates using the polarized fusion image data (fusVH) and the fusion image data (fussym) to which the threshold has been adaptively applied. Claim 2 A method for detecting small fishing vessels according to claim 1, further comprising the step of an average filtering unit (600) adding average filtered image data of the real and imaginary components of the fused image data (fussym) to the fused image data (fussym) to suppress sea clutter. Claim 3 In claim 1, the polarization fusion image data (fusVH) generation step is the VH polarization channel data (S VH ) and the above VV polarization channel data (S VV VH polarization channel strength (|S) from ) VH |²) and VV polarization channel strength (|S VV A step in which |²) is extracted; and the difference between the VV polarization channel strength and the VH polarization channel strength If it is less than 6.53 dB, the value obtained by subtracting 6.53 dB from the above VV polarization channel strength is used, while the difference between the above VV polarization channel strength and the above VH polarization channel strength A method for detecting small fishing vessels, comprising the step of using the above VH polarized channel strength when it is 6.53 dB or higher. Claim 4 A method for detecting small fishing vessels according to claim 1, wherein the step of generating the fused image data (fussym) comprises: a step of extracting a polarized fused image intensity and a cross-correlation image intensity from the polarized fused image data (fusVH) and the cross-correlation image data (sym); and a step of using the larger value (Max) between the cross-correlation image intensity and the polarized fused image intensity when the difference between the cross-correlation image intensity and the polarized fused image intensity is less than 7.2 dB and the cross-correlation image intensity is greater than -18.03 dB, while using the polarized fused image intensity when the difference between the cross-correlation image intensity and the polarized fused image intensity is 7.2 dB or more and the cross-correlation image intensity is -18.03 dB or less. Claim 5 A method for detecting small fishing vessels according to claim 1, wherein the step of determining a candidate for small fishing vessel detection comprises a step of classifying a target as a commercial vessel when the target's RCS is -9.49 dB or higher, and determining the target as a small fishing vessel when the target's RCS is less than -9.49 dB.

Images