Marine radar target detection algorithm and system based on space-time joint filtering technology

Abstract

The invention discloses a marine radar target detection algorithm and system based on a space-time joint filtering technology, and belongs to the technical field of marine remote sensing. In order to solve the problem of low target detection probability under strong sea clutter interference, three-stage processing is adopted: firstly, same-frequency interference suppression is performed on an original radar image sequence, polar coordinate echo data is converted to a frequency wavenumber domain through three-dimensional Fourier transform, and sea clutter energy in a frequency band [kx, ky] is filtered by using a dispersion relation filter; secondly, intercepting a central region of a specific wave number spectrum, extracting target energy meeting a linear equation by adopting a Hough straight line detection filter, and reconstructing a target enhanced image through three-dimensional inverse Fourier transform; and finally, performing target point fine detection on the processing sequence. According to the method, the strong sea clutter signal can be effectively suppressed and the target signal can be extracted only by using the echo amplitude information of the radar, so that the signal-to-noise ratio of the image is enhanced, the target detection probability under the strong sea clutter condition is further improved, and the method is suitable for all marine radars.

Description

Technical Field

[0001] The present invention belongs to the field of marine remote sensing technology, and in particular relates to a marine radar target detection algorithm and system based on spatiotemporal joint filtering technology. Background Art

[0002] Research on marine radar target detection technology in complex sea clutter environments has always been a challenging and hot topic. When a marine radar detects the sea surface, the radar echo contains not only target information but also information about the sea clutter. Because the size and echo intensity of marine targets are relatively small relative to the sea clutter, the target is difficult to detect due to the influence of sea clutter.

[0003] Target detection methods in existing literature are primarily categorized into seven categories: spatial domain processing, time domain processing, frequency domain processing, time-frequency domain processing, and space-time domain processing. Spatial domain processing methods use statistical methods to analyze the mechanisms and characteristics of sea clutter and establish a sea clutter distribution model. Proposed sea clutter distribution models include the lognormal and Weibull distributions based on multi-parameter non-Gaussian models; the K distribution and generalized Pareto (GP) distribution, composite models of texture and speckle components based on a dual-scale model; and the KA distribution, which adds an additional scattering component to match sea spikes. These methods are significantly affected by prior data, resulting in unstable and ineffective detection performance in complex and changing sea clutter environments.

[0004] Time-domain processing methods primarily rely on fractal theory, leveraging the self-similarity and scale invariance of sea clutter echoes to distinguish sea clutter signals from target signals. However, fractal methods cannot capture the state of moving targets and are significantly affected by complex sea clutter. Because the fractal characteristics of sea clutter time series often exist within a time interval, fractal methods require the radar to remain in dwell or slow-scan mode and cannot be applied to radars in rapid scanning modes. Frequency-domain processing methods primarily include moving target indicators (MTIs), moving target detection (MTDs), eigenvalue decomposition (Eigenvalue Decomposition (Singular Value Decomposition (SVDs)), and singular value decomposition (SVDs). MTIs and MTDs exploit the principle that Doppler-shifted moving target frequencies can be separated between multiple consecutive pulses in the frequency domain. By canceling these pulses, the frequency spectrum of the moving target is obtained, enabling sea clutter suppression and target detection. Both EVD and SVD methods are subspace decompositions, based on the first-order Bragg peak characteristics of sea clutter in the frequency domain. These methods require information from multiple coherent pulse echoes of the radar signal. Time-frequency domain processing methods exploit the characteristic that sea clutter and targets cluster in different regions in the time-frequency domain to detect targets. By using different time-frequency transforms and searching for the optimal transform domain to estimate the target's motion parameters, this approach ultimately locates the frequency concentration of the target signal within the time-frequency domain. Methods such as the short-time Fourier transform, fractional Fourier transform, and sparse Fourier transform fall into this category. These methods only have good energy accumulation capabilities for linear frequency modulated (LFM) signals.

[0005] Due to its low cost, ease of operation, all-weather capability, and long range, marine radar has become a crucial electronic device for guiding ships safely and a must-have for all types of vessels. However, the echo signals of rapidly rotating, incoherent oceanographic radars with a single-pulse system contain only amplitude and position information, lacking phase information. Furthermore, pulses in rapid scanning mode cannot be accumulated over long periods of time. Therefore, a spatiotemporal processing method that utilizes only amplitude information to suppress sea clutter and detect targets is suitable for use with these marine radars. Summary of the Invention

[0006] The purpose of the present invention is to provide a marine radar target detection algorithm and system based on spatiotemporal joint filtering technology to solve the influence of strong sea clutter signals in marine radar images on target detection, suppress the strong sea clutter signals in radar images and extract the target signals, thereby improving the target detection probability in the case of strong sea clutter.

[0007] The purpose of the present invention is achieved through the following technical solutions:

[0008] A marine radar target detection algorithm based on spatiotemporal joint filtering technology has the following specific steps:

[0009] Step 1: Use the frequency elimination algorithm to perform frequency interference suppression on the m consecutive original radar image sequences;

[0010] Step 2: Select the area to be processed and interpolate the echo intensity in the polar coordinate system (θ, ρ) to the rectangular coordinate system grid to generate the spatiotemporal image sequence η(x, y, t), where x and y are spatial grid points and t is the number of images in the time series.

[0011] Step 3: Perform a three-dimensional Fourier transform on η(x, y, t) to obtain a three-dimensional frequency wavenumber domain image spectrum I(K x ,K y ,ω);

[0012] Step 4: Combine the frequency domain and wave number domain filters based on the dispersion relation to remove I(K x ,K y ,ω) The sea clutter energy of the frequency band is output, and the filtered image spectrum E(K x ,K y ,ω);

[0013] Step 5: Intercept E(K x ,K y ,ω) in the center area of ​​the a-th and b-th wavenumber spectra;

[0014] Step 6: Use the frequency domain and wave number domain joint filter based on line detection to obtain the value of E(K x ,K y ,ω) to extract the linear equation k y,ω =K r ·k x,ω +b ω,r The target energy is obtained by extracting the three-dimensional frequency wave number image spectrum M(k x,ω ,k y,ω ,ω);

[0015] Step 7: For M(k x,ω ,k y,ω ,ω) performs three-dimensional inverse Fourier transform and outputs the radar image sequence η after suppressing the sea clutter signal m (x,y,t);

[0016] Step 8: η m (x, y, t) to detect target points, thereby improving the target detection probability in strong sea clutter conditions.

[0017] Furthermore, the formula for the three-dimensional Fourier transform in step 3 is:

[0018]

[0019] Among them, L x , L y are the scales of the image sequence in the horizontal and vertical directions in the spatial domain respectively; T is the scale of the image sequence in the time domain; k x , k y are the horizontal and vertical wavenumber components of the image spectrum in the wavenumber domain, and ω is the frequency of the image spectrum in the frequency domain.

[0020] Furthermore, the step 4 is to I(K x ,K y ,ω) is filtered as follows:

[0021]

[0022] Among them, K p is the upper boundary of the filter band, K n is the lower band boundary of the filter.

[0023] Furthermore, the step 5 intercepts E(K x ,K y ,ω) in the center of the a-th and b-th wavenumber spectra is and Where i is the scale of the intercepted central area.

[0024] Furthermore, in step 6, bω,r The calculation formula is:

[0025]

[0026] Among them, K r is the slope vector of all straight lines detected by the straight line detection algorithm, b ω,r are all the line intercept vectors detected by the line detection algorithm, n is the serial number of each wavenumber spectrum in the wavenumber frequency domain, C i is the width expansion coefficient vector, and r is the serial number of all straight lines in the image.

[0027] Furthermore, the three-dimensional inverse Fourier transform in step 7 is:

[0028]

[0029] Where t is the number of images in the time domain.

[0030] Furthermore, the method only utilizes the radar echo amplitude information and does not rely on the Doppler information of the echo or long-term accumulation of pulse information.

[0031] A computer device / equipment / system comprising a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of a marine radar target detection algorithm based on a spatiotemporal joint filtering technique.

[0032] The beneficial effects of the present invention are:

[0033] The present invention first removes sea clutter energy using a combined time-domain and spatial-domain filter. Second, it extracts target energy using this combined time-domain and spatial-domain filter, completing the coarse detection phase. Third, it uses a detector to detect target points, completing the fine detection phase. Compared to existing technologies, the present invention only utilizes radar echo amplitude information, eliminating the need for echo Doppler information or long-term pulse accumulation. This effectively suppresses strong sea clutter signals and extracts target signals, thereby enhancing the image's signal-to-noise ratio and, in turn, improving the probability of target detection in strong sea clutter conditions. The invention is applicable to all marine radars. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is a single original radar image;

[0035] Figure 2 A single radar image after interpolation for the selected area;

[0036] Figure 3 is the two-dimensional cross section of the three-dimensional frequency-wavenumber domain image spectrum along the |K| direction;

[0037] Figure 4 is the two-dimensional cross section along the |K| direction of the three-dimensional frequency-wavenumber domain image spectrum after filtering out sea clutter;

[0038] Figure 5 It is the two-dimensional cross section along the |K| direction of the three-dimensional frequency-wavenumber domain image spectrum after the target is extracted;

[0039] Figure 6 A single radar image after target extraction;

[0040] Figure 7 is the target detection result of a single radar image;

[0041] Figure 8 Flowchart of the implementation method. DETAILED DESCRIPTION

[0042] The present invention will be further described below with reference to the accompanying drawings.

[0043] The present invention provides a marine radar target detection algorithm and system based on spatiotemporal joint filtering technology. The specific steps of the algorithm are as follows:

[0044] Step 1: Perform sea clutter suppression on the original radar image sequence of the selected area. The area in the original radar image sequence where sea clutter is to be suppressed is selected, and a three-dimensional frequency-wavenumber domain image spectrum is obtained using a three-dimensional Fourier transform (3D-FFT). A spatiotemporal joint filter based on the dispersion relation is selected to remove sea clutter energy.

[0045] Step 1.1: Use the selected frequency elimination algorithm to perform frequency interference suppression on the continuous m original radar image sequence.

[0046] Step 1.2: Select the region in the original radar image sequence obtained in step 1.1 where sea clutter needs to be suppressed. Using the selected interpolation algorithm, interpolate the echo intensities of the points within the selected region onto a rectangular coordinate grid. The center of this grid has an angle θ in the polar coordinate system, a radial distance ρ, and a size of x × y. This yields the echo intensity sequence η(x, y, t), where x is the number of points in the horizontal direction, y is the number of points in the vertical direction, and t is the number of image frames in the time series.

[0047] Step 1.3, perform a three-dimensional Fourier transform on the image sequence η(x, y, t) obtained in step 1.2 to obtain a three-dimensional frequency wavenumber domain image spectrum I(k x ,k y ,ω).

[0048]

[0049] Where: L x , L yare the horizontal and vertical scales of the image sequence in the spatial domain, respectively, and T is the scale of the image sequence in the temporal domain. x , k y are the horizontal and vertical wavenumber components of the image spectrum in the wavenumber domain, and ω is the frequency of the image spectrum in the frequency domain.

[0050] Step 1.4, select a frequency domain and wave number domain joint filter based on dispersion relation to filter the three-dimensional frequency wave number image spectrum I(K x ,K y ,ω) filtering. The three-dimensional frequency wavenumber image spectrum E(K x ,K y ,ω).

[0051]

[0052] Among them: K p is the upper boundary of the filter band, K n is the lower boundary of the filter band, ω is the frequency of the image spectrum in the frequency domain, k x , k y are the horizontal and vertical wavenumber components of the image spectrum in the wavenumber domain, respectively.

[0053] Step 2: Target extraction is performed on the original radar image sequence after sea clutter suppression. A spatiotemporal joint filter based on the Hough line detection algorithm is selected to extract target energy. A three-dimensional inverse Fourier transform (3D-IFFT) is performed on the three-dimensional frequency-wavenumber domain image spectrum of the filtered image sequence to obtain the image sequence after target extraction.

[0054] Step 2.1, intercept the three-dimensional frequency wave number image spectrum E(k x ,k y ,ω) in the two wave number spectra E(k x ,k y ,a) and E(k x ,k y ,b) the central area on and a and b are the ath and bth image spectra in the selected three-dimensional frequency wavenumber image spectrum, respectively. i is the scale of the intercepted central area.

[0055] Step 2.2, the and A frequency domain and wavenumber domain joint filter based on line detection is used to analyze the three-dimensional frequency wavenumber image spectrum E(K x ,K y ,ω) filtering to obtain the three-dimensional frequency wave number image spectrum M(k x,ω ,ky,ω ,ω).

[0056]

[0057] Among them: K r is the slope vector of all straight lines detected by the straight line detection algorithm, b ω,r are all the line intercept vectors detected by the line detection algorithm, n is the serial number of each wavenumber spectrum in the wavenumber frequency domain, a and b are the serial numbers of the two selected Hough detection images, C i is the width expansion coefficient vector, and r is the serial number of all straight lines in the image.

[0058] Step 2.3, the three-dimensional frequency wave number image spectrum M(k x ,k y ,ω) to perform three-dimensional inverse Fourier transform and obtain the radar image sequence η after suppressing the sea clutter signal m (x,y,t).

[0059]

[0060] Where: L x , L y K is the scale of the image sequence in the horizontal and vertical directions in the spatial domain, and T is the scale of the image sequence in the time domain. x , K y where x and y are the horizontal and vertical wavenumber components of the image spectrum in the wavenumber domain, respectively; ω is the frequency of the image spectrum in the frequency domain; x and y are the number of points in the horizontal and vertical directions in the spatial domain, respectively; and t is the number of image frames in the time domain.

[0061] Step 3: Perform target point precision detection on the processed image sequence.

[0062] A detector is used to detect target points in the image sequence after target extraction obtained in step 2.3, thereby improving the target detection probability in the case of strong sea clutter.

[0063] Example 1:

[0064] The marine radar used in the embodiments of the present invention is an X-band navigation radar operating in short pulse mode. After digitization, the echo data is stored in polar coordinates by line. The time interval between two adjacent stored lines is less than 1 ms. The radar antenna scans one circle in approximately 2.5 seconds. A radar image has 2048 lines, each line has 2048 pixels, and its azimuth resolution is approximately 0.1°, and its radial resolution is approximately 2.5 m.

[0065] The main technical parameters of the marine radar are shown in Table 1:

[0066] Table 1 Technical parameters of marine radar

[0067]

[0068] according to Figure 1-8 , the specific steps of the experiment are:

[0069] The first step is to suppress sea clutter in the original radar image sequence of the selected area. This includes the following steps:

[0070] Step 1.1, obtain a continuous sequence of 32 original radar images, attached Figure 1 A single original radar image is used. Spatial domain correlation is used to suppress co-channel interference on the radar image. Specifically, in areas close to the radar antenna, the echo intensity of each pixel is replaced by the median of the echo intensities of two non-noise points on the left and right adjacent storage lines with the same radial distance from the pixel. In areas farther from the radar antenna, the echo intensity of each pixel is replaced by the median of the echo intensities of seven non-noise points within a 3×3 neighborhood window.

[0071] Step 1.2: Select the area in the original radar image sequence obtained in step 1.1 where sea clutter needs to be suppressed. Use the nearest point interpolation algorithm to interpolate the echo intensities of the points in the selected area into a rectangular coordinate grid. The center point of the grid has an angle of 0° in the polar coordinate system, a radial distance of 0m, and a size of 848×848. The resulting echo intensity sequence η has a dimension of 848×848×32. 848 is the number of points in the horizontal and vertical directions, and 32 is the number of amplitudes in the time series. Figure 2 A single radar image after interpolation of the selected area.

[0072] Step 1.3, perform a three-dimensional Fourier transform on the image sequence η obtained in step 1.2 to obtain a three-dimensional frequency wavenumber domain image spectrum I(k x ,k y ,ω), attached Figure 3 is the two-dimensional cross section of the three-dimensional frequency-wavenumber domain image spectrum along the |K| direction.

[0073]

[0074] Where: L x , L y k is the scale of the image sequence in the horizontal and vertical directions in the spatial domain, which is 6,360m, and T is the scale of the image sequence in the time domain, which is 80s. x , k y are the horizontal and vertical wavenumber components of the image spectrum in the wavenumber domain, ω is the frequency of the image spectrum in the frequency domain, I(k x ,ky ,ω) has a dimension of 848×848×32.

[0075] Step 1.4, the three-dimensional frequency wave number image spectrum I(k x ,k y ,ω) is converted into a three-dimensional wavenumber frequency domain image spectrum I(|K|,θ1,ω) in the polar coordinate system

[0076] I(|K|,θ1,ω)=I(k x ,k y ,ω)·β

[0077] Where |K| is the wave number modulus, θ1 is the image spectrum angle, and β is the conversion matrix from rectangular coordinate system to polar coordinate system.

[0078] Step 1.5: Integrate I(|K|,θ1,ω) by wavenumber angle to obtain the image spectrum I(|K|,ω) in the two-dimensional wavenumber mode frequency domain.

[0079]

[0080] Where: |K| is the wave number mode; θ1 is the energy spectrum angle, and ω is the frequency of the image spectrum in the frequency domain.

[0081] In step 1.6, the maximum spectral value in the two-dimensional energy spectrum I(|K|,ω) is calculated to be 11829237.6. The wavenumber modulus values ​​of all points with spectral values ​​greater than 11237775.72 are obtained. The average of the obtained wavenumber modulus values ​​is 0.0487.

[0082] Step 1.7, the three-dimensional frequency wave number image spectrum I(k x ,k y ,ω) filtering. Figure 4 is the two-dimensional cross section of the three-dimensional frequency-wavenumber domain image spectrum along the |K| direction after filtering out sea clutter.

[0083]

[0084]

[0085] Among them: K p is the upper boundary of the filter band, K n is the lower band boundary of the filter, ω is the frequency of the energy spectrum in the frequency domain, Δω is the frequency resolution 0.0785, and g is 9.8 m / s 2 , ΔK is the wave number mode resolution 0.0065, U max 3m / s.

[0086] The second step is to extract targets from the original radar image sequence after sea clutter suppression. This includes the following steps:

[0087] Step 2.1, intercept the three-dimensional frequency wave number image spectrum E(k x ,k y ,ω) in the central area of ​​the 16th and 18th frames and

[0088] Step 2.2: Perform Hough line detection on the image spectrum obtained in step 2.1. Detect the slope vector K in the two images. r is [0°, 90°, 45°], the intercept vector b ω,r is a 2×3 matrix

[0089] Step 2.3, the three-dimensional frequency wave number image spectrum E(k x ,k y ,ω) filtering to obtain the three-dimensional frequency wave number image spectrum M(k x,ω ,k y,ω ,ω).

[0090]

[0091] Among them: K r is the slope vector of all straight lines detected by Hough, b ω,r are all the straight line intercept vectors in the two images detected by Hough, n is the serial number of each wavenumber spectrum in the wavenumber frequency domain, a and b are the serial numbers 16 and 18 of the two selected Hough detection images, C i The width expansion coefficient vector value is 10, and r is the serial number of all straight lines in the image, which is 1 to 3.

[0092] Step 2.4, perform a three-dimensional inverse Fourier transform on the three-dimensional frequency wavenumber image spectrum obtained after extracting the target energy in step 2.3 to obtain the radar image sequence η after suppressing the sea clutter signal m (x,y,t)

[0093]

[0094] The third step is to detect target points in the image sequence after suppressing sea clutter. This includes the following steps:

[0095] The CA-CFAR detector is used to detect target points on the last image in the image sequence after target extraction. When the first image is removed from the 32 radar image sequences after target extraction and a new image is added after the last image, target point detection is performed on the latest image in the updated radar image sequence. Figure 6 Target detection results for a single radar image.

[0096] A marine radar target detection algorithm based on spatiotemporal joint filtering technology, proposed in this paper, was applied to a large amount of radar data acquired by a test vessel during navigation in the East China Sea in 2017, along with sea condition information from the relevant time periods. This experiment selected data from October 20, 22, and 23, 2017, encompassing three types of sea conditions: low, medium, and high. Comparative experiments on sea clutter suppression and target detection were conducted on the original radar images using both the proposed method and the EMD method to verify the performance of the proposed method.

[0097] The performance of the two methods is compared using the signal-to-noise ratio and detection probability, which are calculated as follows:

[0098]

[0099] Where s and x are the sum of the echo intensities of all target points and all clutter points, respectively. <·> is the average value. s(n) and x(n) are the sum of the echo intensities of the target points and the sea clutter points within the range selected in this method, respectively.

[0100]

[0101] Among them, N t is the total number of target points, N dt is the total number of detected target points.

[0102] This study experimentally analyzed a large amount of radar data and sea condition information collected by a test vessel during its voyage in the East China Sea in 2017. Data from October 20, 22, and 23, 2017, covering three sea conditions: low, medium, and high. Comparative experiments on sea clutter suppression and target detection were conducted on the original radar images using the method presented in this study and the EMD method, respectively, to verify the performance of the present invention.

[0103] The comparison results are shown in Table 2, Table 3, Table 4, and Table 5:

[0104] Table 2 Effects of two methods on image signal-to-noise ratio improvement under low sea conditions

[0105]

[0106] Table 3 Effects of two methods on image signal-to-noise ratio improvement under sea conditions

[0107]

[0108] Table 4 Effects of two methods on image signal-to-noise ratio improvement under high sea conditions

[0109]

[0110] Table 5 Detection results of two detectors

[0111]

[0112] Example 2:

[0113] All steps in the first and second steps are exactly the same as those in the first and second steps in Example 1.

[0114] The third step is to perform target point detection on the image sequence after target extraction. A Weibull-CFAR detector is used to detect target points on the last image in the sea clutter suppressed image sequence. After removing the first image from the 32 radar image sequences with sea clutter suppressed and adding a new image after the last, target point detection is performed on the latest image in the updated radar image sequence.

[0115] Example 3:

[0116] All steps in the first and second steps are exactly the same as those in the first and second steps in Example 1.

[0117] The third step is to perform target point detection on the image sequence after target extraction. A support vector machine (SVM) detector is used to detect target points on the last image in the sea clutter suppressed image sequence. After removing the first image from the 32 radar image sequences with sea clutter suppressed and adding a new image after the last, target point detection is performed on the latest image in the updated radar image sequence.

[0118] Experimental results show that the algorithm proposed in this invention can effectively suppress strong sea clutter signals and extract target signals, thereby improving the target detection probability in strong sea clutter conditions.

[0119] The marine radar target detection algorithm based on spatiotemporal joint filtering technology proposed in the present invention solves the influence of strong sea clutter signals on the detection target in marine radar image target detection.

[0120] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A marine radar target detection algorithm based on spatiotemporal joint filtering technology, characterized by: The specific steps are as follows: Step 1: Use the frequency elimination algorithm to perform frequency interference suppression on the m consecutive original radar image sequences; Step 2: Select the area to be processed and interpolate the echo intensity in the polar coordinate system (θ, ρ) to the rectangular coordinate system grid to generate the spatiotemporal image sequence η(x, y, t), where x and y are spatial grid points and t is the number of images in the time series. Step 3: Perform a three-dimensional Fourier transform on η(x, y, t) to obtain a three-dimensional frequency wavenumber domain image spectrum I(K x ,K y ,ω); Step 4: Combine the frequency domain and wave number domain filters based on the dispersion relation to remove I(K x ,K y ,ω) The sea clutter energy of the frequency band is output, and the filtered image spectrum E(K x ,K y ,ω); Step 5: Intercept E(K x ,K y ,ω) in the center area of ​​the a-th and b-th wavenumber spectra; Step 6: Use the frequency domain and wave number domain joint filter based on line detection to obtain the value of E(K x ,K y ,ω) to extract the linear equation k y,ω =K r ·k x,ω +b ω,r The target energy is obtained by extracting the three-dimensional frequency wave number image spectrum M(k x,ω ,k y,ω ,ω); Step 7: For M(k x,ω ,k y,ω ,ω) performs three-dimensional inverse Fourier transform and outputs the radar image sequence η after suppressing the sea clutter signal m (x,y,t); Step 8: η m (x, y, t) to detect target points, thereby improving the target detection probability in strong sea clutter conditions.

2. The marine radar target detection algorithm based on spatiotemporal joint filtering technology according to claim 1, characterized in that: The formula for the three-dimensional Fourier transform in step 3 is: Among them, L x , L y are the scales of the image sequence in the horizontal and vertical directions in the spatial domain respectively; T is the scale of the image sequence in the time domain; k x , k y are the horizontal and vertical wavenumber components of the image spectrum in the wavenumber domain, and ω is the frequency of the image spectrum in the frequency domain.

3. The marine radar target detection algorithm based on spatiotemporal joint filtering technology according to claim 1 is characterized by: The step 4 is to x ,K y ,ω) is filtered as follows: Among them, K p is the upper boundary of the filter band, K n is the lower band boundary of the filter.

4. The marine radar target detection algorithm based on spatiotemporal joint filtering technology according to claim 1, characterized in that: The step 5 intercepts E(K x ,K y ,ω) in the center of the a-th and b-th wavenumber spectra is and Where i is the scale of the intercepted central area.

5. The marine radar target detection algorithm based on spatiotemporal joint filtering technology according to claim 1 is characterized in that: Step 6 b ω,r The calculation formula is: Among them, K r is the slope vector of all straight lines detected by the straight line detection algorithm, b ω,r are all the line intercept vectors detected by the line detection algorithm, n is the serial number of each wavenumber spectrum in the wavenumber frequency domain, C i is the width expansion coefficient vector, and r is the serial number of all straight lines in the image.

6. The marine radar target detection algorithm based on spatiotemporal joint filtering technology according to claim 1, characterized in that: The three-dimensional inverse Fourier transform in step 7 is: Where t is the number of images in the time domain.

7. The marine radar target detection algorithm based on spatiotemporal joint filtering technology according to claim 1 is characterized in that: The method only utilizes the radar echo amplitude information and does not rely on the Doppler information of the echo or long-term accumulation of pulse information.

8. A computer device / apparatus / system comprising a memory, a processor, and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

Images