An underwater target detection device and its operation method based on forward-looking sonar partitioning

Abstract

This invention discloses an underwater target detection device and operating method based on forward-looking sonar partitioning, relating to the field of underwater acoustic signal processing technology. This invention, by displaying the measurement range, divides the beam domain into grids from two dimensions: beam and range, forming multiple local detection grids. It compresses and images the corresponding beam domain data to form image domain data, and determines the detection threshold for each local detection grid based on the signal-to-noise ratio of the corresponding image domain data to identify underwater targets. By dividing the beam domain data of the detection area into grids according to beam and range, this invention can reasonably divide the beam domain data generated by underwater targets, underwater reverberation, and acoustic ray changes in the beam domain data corresponding to the signal domain. It compresses and images the beam domain data of each local detection grid, and accurately identifies the parameters corresponding to underwater targets, underwater reverberation, and acoustic ray changes from the perspective of obtaining the detection threshold from the signal-to-noise ratio of the image domain data, thus accurately identifying underwater targets.

Description

Technical Field

[0001] This invention relates to the field of underwater acoustic signal processing technology, and in particular to an underwater target detection device and its operation method based on forward-looking sonar partitioning. Background Technology

[0002] Forward-looking sonar is a device used for underwater target detection, calibration, tracking, and identification. During underwater target detection using forward-looking sonar, the signal is transmitted into the water through a power amplifier and transducer, illuminating the target being detected and reflecting back to the receiving hydrophone. The signal received by the hydrophone contains not only the target echo but also reverberation and noise interference. The interference experienced by forward-looking sonar mainly comes from the unevenness of the underwater topography and the scattering effect of various scatterers on the seabed, resulting in reverberation caused by the scattering of sound waves propagating to the seabed. In addition, under the combined influence of temperature, pressure, and salinity, the speed of sound propagation in open sea water corresponds to the water depth. The difference in propagation speed leads to the refraction of underwater acoustic signals, that is, the transmission path bends in the direction of lower sound speed.

[0003] For underwater target detection, the sonar system has known transmitted information and needs to determine the target information, which is contained in the echo signal reflected back to the receiver. Target detection is generally performed in the signal domain. By analyzing the echo signal and comparing it with the transmitted signal parameters, the target information contained in the echo (such as position and velocity) can be obtained. Therefore, the sonar system needs to consider more complex issues, including: the target information the receiver needs to obtain comes from the response excited by the transmitted information, but the transmitted waveform needs to be within the allowable range of the transmission channel, so the transmitted waveform needs to be designed to increase the target information contained in the echo; the information that active sonar needs to extract from the target echo is actually the posterior probability P(x|Y), that is, extracting the information of x from the received waveform Y, but it also faces interference from reverberation and background noise, so it is necessary to consider how to perform detection filtering on waveform Y to reduce interference terms; the target features may have N parameters, and different combinations of parameters correspond to different patterns, so the sonar information processing process can be divided into two main parts: one is the feature extraction of the target, and the other is the pattern recognition processing of the target.

[0004] Therefore, when using the received signal domain data for target detection, the signal domain data contains a large number of signals due to underwater reverberation and changes in sound ray. When analyzing the signal domain data, the parameters corresponding to the target's features are mixed with the parameters corresponding to underwater reverberation and changes in sound ray, making it difficult to identify them and thus making it difficult to accurately detect underwater targets. Summary of the Invention

[0005] This invention provides an underwater target detection device and operating method based on forward-looking sonar partitioning, which can solve the problem in the prior art where the parameters corresponding to the characteristics of the target are mixed with the parameters corresponding to underwater reverberation and sound ray changes when analyzing signal domain data, making it difficult to identify and thus difficult to accurately detect underwater targets.

[0006] This invention provides an underwater target detection device based on forward-looking sonar partitioning, including a beam domain data module, a partitioned detection grid control module, a compression module, a partitioned target detection module, and a target detection result aggregation module;

[0007] The beam domain data module is used to identify parameters of the beam domain data from the signal domain processor in order to form forward-looking sonar beam domain data.

[0008] The partitioned detection grid control module divides the beam domain data into grids based on the current detection range of the forward-looking sonar, from two dimensions: beam and range, to form a local detection grid set. The local detection grid is formed by the azimuth angle between the sonar horizontal plane and the sonar to the target and the detection distance between the sonar and the target, thus forming a local detection area.

[0009] The compression module is used to perform one-dimensional convolution on the beam domain data in each local detection grid, take the maximum value, compress and image it to form the image domain data of each local detection grid.

[0010] The partitioned target detection module obtains the signal-to-noise ratio within each local detection grid based on the forward-looking sonar active acoustic reflection signal and the environmental noise level of the image domain data corresponding to each local detection grid, and determines the detection threshold of each local detection grid. Based on the detection threshold and the corresponding azimuth angle and detection distance, it obtains multiple underwater targets within each local detection area.

[0011] The target detection result aggregation module is used to determine whether multiple underwater targets within each local detection grid have an intersection in the target location range, and merges the positions of the intersecting targets to form an underwater target.

[0012] Preferably, the beam domain data is data containing target direction, sonar weight information, sonar frequency information and sonar timing information after sonar beamforming;

[0013] The image domain data is obtained by compressing the beam domain data and converting it into eight-bit image data.

[0014] Preferably, the current detection range of the forward-looking sonar is segmented in 100-meter segments.

[0015] The local detection grid uses 32 as an indicator, meaning that the detection grid is formed by the detection area covered by 32 beams, and each local detection grid corresponds to a 50-meter range.

[0016] Preferably, the compression module performs one-dimensional convolution on the beam domain data within each local detection grid, takes the maximum value, and then compresses the data to a size of 512*512.

[0017] Preferably, obtaining the detection threshold for each local detection region includes:

[0018] The beam domain data corresponding to each local detection grid is compressed and stored in the scanDetect16Buffer buffer;

[0019] Obtain the mean and standard deviation (std_dev) of the beam domain data corresponding to each local detection grid within the scanDetect16Buffer. Then, the grid background for each local detection grid is:

[0020] background=mean+2*std_dev;

[0021] Based on the background grid, the detection threshold for each local detection grid is obtained, then:

[0022] threshold=background / db

[0023] Where: threshold is the detection threshold; db is the signal-to-noise ratio.

[0024] Preferably, the step of merging the positions of overlapping targets includes:

[0025] Obtain the target contour area of ​​multiple underwater targets within each local detection region, and discard contours with s>400 and s<3;

[0026] For a target profile that meets the conditions, calculate its centroid coordinates mX and mY, and convert the centroid coordinates into sampling point mBin and beam mBeam;

[0027] The coordinate error offset range of each centroid is 10 meters. The two centroids are compared. If there is an intersection, they are merged. The merged centroid is the midpoint between the original two centroids.

[0028] This invention also provides an operation method for an underwater target detection device based on forward-looking sonar partitioning, comprising the following steps:

[0029] The beam domain data module receives the range and beam domain data parameters displayed on the display console and stores them in a shared buffer pool;

[0030] The partition detection grid control module determines the partition detection grid based on the display range and the forward sonar opening angle, and assigns the grid parameters to the corresponding partition detection task processing thread;

[0031] The partition target detection module extracts the corresponding beam domain data from the system data buffer and performs compressed imaging on the beam domain data; the partition target detection module determines the detection threshold based on the signal-to-noise ratio of the detection area, performs multi-target detection in the regional image domain data, and merges similar targets;

[0032] The target detection results summary module summarizes the targets detected by the target detection module in each partition.

[0033] Preferably, the number of partition detection task processing threads is set according to the full scale and the maximum opening angle of the sonar, and each thread deploys one partition detection task thread.

[0034] This invention provides an underwater target detection device and operating method based on forward-looking sonar partitioning. Compared with the prior art, its advantages are as follows:

[0035] This invention utilizes forward-looking sonar beam domain data and the current detection range of the forward-looking sonar to divide the beam domain data into multiple local detection grids based on both beam and range dimensions. The beam domain data within each local detection grid is compressed and imaged to form image domain data. The detection threshold for each local detection grid is determined based on the signal-to-noise ratio (SNR) of the corresponding image domain data, thereby identifying underwater targets. In other words, by dividing the beam domain data of the detection area into grids according to beam and range, this invention can reasonably divide the beam domain data generated by underwater targets, underwater reverberation, and acoustic variations within the beam domain data corresponding to the signal domain. Then, the beam domain data of each local detection grid is compressed and imaged. By obtaining the detection threshold from the SNR of the image domain data, the parameters corresponding to underwater targets, underwater reverberation, and acoustic variations are accurately identified, thus accurately identifying underwater targets.

[0036] Furthermore, when detecting underwater multiple targets in the image domain, this invention can acquire features such as the shape and shadow of underwater targets, which is more intuitive and accurate than detection in the signal domain.

[0037] Moreover, the detection threshold obtained by each local detection grid in this invention is dynamic, which greatly improves the detection rate when the signal of distant targets is weak. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the overall architecture of an underwater target detection device and operation method based on forward-looking sonar partitioning provided in an embodiment of the present invention;

[0039] Figure 2 This is a schematic diagram of a multi-grid area target detection method based on a forward-looking sonar partitioning underwater target detection device and operation method provided in an embodiment of the present invention.

[0040] Among them: 101, beam domain data module; 102, partitioned detection grid control module; 103, compression module; 104, partitioned target detection module; 105, target detection result summary module. Detailed Implementation

[0041] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention can be practiced in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0042] See Figures 1-2 This invention provides an underwater target detection device and operation method based on forward-looking sonar partitioning, including a host system. The host system includes: a beam domain data module 101, a partitioned detection grid control module 102, a compression module 103, a partitioned target detection module 104, and a target detection result aggregation module 105.

[0043] Compression module 103: Performs one-dimensional convolution maximum compression on the beam domain data to form image domain data.

[0044] Partition Detection Grid Control Module 102: Based on the displayed range, it partitions the beam domain data within the range to form a detection grid.

[0045] Partition Target Detection Module 104: Performs underwater target detection in the detection task partition and fuses the data based on the target location.

[0046] Target Detection Result Summary Module 105: Summarizes the target detection results in the detection zone.

[0047] The host system consists of multiple DSPs. Specifically: forward-looking sonar beam domain data and beam domain data parameters are uploaded to the host system from the signal domain processor; a thread is allocated to each zone to run the zone detection module; the zone detection grid control module 102 divides the grid according to the range displayed on the control console, from the two dimensions of beam and distance, to form a local detection grid set; the local detection grid forms a local detection area by the azimuth angle and the detection distance; the zone target detection module 104 corresponds to the zone grid and is used to perform target detection within the local grid area.

[0048] RAW data consists of 16-bit data generated after sonar beamforming, while image domain data is generated by compressing beam domain data and converting it into 8-bit bitmap data.

[0049] Among them, the detection area grid division is to divide the beam domain data and perform target detection in the image domain.

[0050] The data within the range of the forward-looking sonar display is dynamically divided into several local detection grids, and multi-target detection is performed on the image domain data corresponding to the local grids. Each grid corresponds to a target detection task thread.

[0051] The host system allocates a DSP core to each partition detection task thread and detection and foreign aggregation thread; the number of partition detection task threads started by the host system is set according to the maximum opening angle of the full-scale sonar core, and each thread deploys a partition detection task module; compression is to perform one-dimensional beam convolution compression and imaging on the partition beam domain data to form image domain data; the partition target detection module 104 performs partition target detection in the image domain, and the detection threshold is determined by the signal-to-noise ratio of the corresponding data in the detection area; for multiple underwater targets detected, the partition detection module merges the positions of targets with overlapping azimuths based on their position errors.

[0052] Target detection occurs in a local area within the range of forward-looking sonar.

[0053] The local target detection module evaluates and determines the detection threshold based on the forward-looking sonar active acoustic reflection signal and the ambient noise level of the local area.

[0054] The data within the range of the forward-looking sonar display is dynamically divided into several local detection grids, with each grid corresponding to a target detection task thread. This process includes the following steps:

[0055] Step 1: Start the host system and start the partition detection task thread and detection summary thread at full capacity.

[0056] Step 2: The host system receives the range displayed on the display console.

[0057] Step 3: The host system receives beam domain data parameters.

[0058] Step 4: The host system receives the beam domain data corresponding to the display range and stores it in the shared buffer pool.

[0059] Step 5: The partition detection grid control module 102 determines the partition detection grid based on the display range and the forward-looking sonar detection opening angle, and assigns the grid parameters to the corresponding detection task processing thread.

[0060] Step 6: The partition target detection module 104 extracts the corresponding beam domain data from the host system data buffer.

[0061] Step 7: The partitioned target detection module 104 performs compressed imaging on the beam domain data of the corresponding detection grid area.

[0062] Step 8: The partition target detection module 104 determines the detection threshold based on the signal-to-noise ratio of the detection area, performs multi-target detection in the regional image domain data, and merges similar targets.

[0063] Step 9: Target detection result summary module 105 summarizes the targets detected by the partition target detection module 104.

[0064] Step 10: The target detection result summarization module 105 determines whether the summarization is complete. If yes, it returns to Step 2; otherwise, it synchronously waits for the parallel partition detection thread in Step 10.

[0065] In step five, the number of partition detection task threads started by the host system is set according to the full scale and the maximum opening angle of the sonar, and each thread deploys one partition detection task module.

[0066] In step seven, the partitioned beam domain data is compressed by one-dimensional beam convolution and imaged to form image domain data.

[0067] In step eight, the partitioned target detection module 104 performs partitioned target detection in the image domain, and the detection threshold is determined by the signal-to-noise ratio of the data corresponding to the detection region.

[0068] In step eight, the partition target detection module 104 merges the positions of multiple detected underwater targets based on the overlapping target position error ranges.

[0069] Specifically:

[0070] After the host system starts up, it receives beam domain data and fills it into the beam domain data buffer rawBuffer. The rawBuffer stores the complete data of the current frame. The data source for multi-target detection is the current frame data stored in the rawBuffer. Target detection is performed on the current frame to form the detection result.

[0071] Define a beam domain data buffer for multi-core parallel compression:

[0072] oneCore16UC1Buffer[512*1024].

[0073] Define a private image domain data buffer for a single DSP core for multi-core parallel compression:

[0074] scanOneCore16UC1Buffer[512*1024].

[0075] Define a detection result data buffer to store single-core detected targets:

[0076] scanDetect16Buffer[512*512].

[0077] Define a single-core detection result buffer:

[0078] gaussiansBuff[256*500].

[0079] Define a multi-core detection result aggregation buffer:

[0080] gaussianAccuBuff[256*16*500].

[0081] Define the detection grid:

[0082] detectGrids[16*16].

[0083] I. System Initialization: Due to the large amount of data in one data frame period of sonar, as in this embodiment, it consists of 512 beams; this embodiment uses the RAPIDIO protocol to transmit data between multiple DSP cores; the system initialization steps are as follows:

[0084] ① Create a receiving beam domain data thread on the host side and deploy it on DSP1 core 0.

[0085] ② Create six beam domain data compression threads on the host side and deploy them on DSP2 cores 0-7.

[0086] ③ Create sixteen target detection threads on the host side and deploy them on DSP3 cores 0-7 and DSP4 cores 0-7.

[0087] ④ Create a target detection result aggregation thread on the host side and deploy it to DSP1 core 1.

[0088] II. Event Handling Flow: This mainly involves receiving beam domain data and performing multi-core parallel detection on the corresponding area. The processing steps are as follows:

[0089] A: The partition detection grid control module 102 reads the display range sent by the display console and divides the range into 100-meter intervals.

[0090] Number of segments divs = range / 100.

[0091] The grid beam rows = 32.

[0092] Where divs is the number of segments, that is, each detection grid corresponds to a 50-meter range, and the detection area covered by 32 beams forms a detection grid, which is filled into detectGrids; since sixteen target detection threads are initialized, the detection method supports a maximum detection distance of 1600.

[0093] B: For the detection grid detectGrids[i], the beam domain data module 101 obtains the corresponding beam domain data in the rawBuffer; the obtained beam domain data corresponding to the current grid is stored in oneCore16UC1Buffer, where oneCore16UC1Buffer is private within the kernel and not shared; since the beam domain data is large, it cannot be directly used for target detection. In this embodiment, the beam domain data is first compressed, and then the compressed data is used for target detection; the oneCore16UC1Buffer data is compressed, and the compression algorithm is shown in Table 1 and Table 2. The compressed result is stored in oneCore16UC1Buffer.

[0094] Table 1 One-dimensional maximum convolution compression algorithm

[0095]

[0096] Table 2 Multi-grid partitioning parallel compression algorithm

[0097]

[0098]

[0099] C: For each detection grid, step B has compressed its data; the beam domain data used by each detection grid is 512*512 after compression and interpolation, and is kept in the scanDetect16Buffer buffer; the scanDetect16Buffer buffer data is subjected to maximum median filtering and 3x3 convolution kernel to suppress noise, with the aim of highlighting the target.

[0100] D: Calculate the mean and standard deviation (std_dev) of scanDetect16Buffer, and estimate the background:

[0101] background=mean+2*std_dev.

[0102] E: Calculate the local detection grid detection threshold based on the estimated background, where the signal-to-noise ratio (SNR) in dB is set by the display console, i.e., how many dB times the background is required to be detected.

[0103] threshold = background / db.

[0104] The values ​​of db are 0.7943, 0.6310, 0.5012, 0.3981, 0.3162, 0.2512, 0.1995, 0.1585, 0.1259, and 0.1000.

[0105] F: Use threshold to binarize the scanDetect16Buffer data and extract the image contour of each suspected target.

[0106] G: Calculate the area s of each suspected target contour, and discard contours with s>400 and s<3.

[0107] H: For each suspected target contour that meets the conditions, calculate its centroid coordinates mX and mY, and convert the centroid coordinates into sampling point mBin and beam mBeam; the error offset range of each centroid coordinate is 10 meters. Compare the two centroids. If there is an intersection, merge them. The merged centroid is the midpoint of the original two centroids.

[0108] I: In each detection kernel, the detection thread saves the results of suspected target contours that meet the conditions to the gaussiansBuff buffer and sends them to the aggregation thread, which then aggregates them. The aggregation module saves the aggregated results to the gaussianAccuBuff buffer.

[0109] like Figure 1 When the partition detection grid control module 102 shown in label 1 is executed, it divides the display range into segments and processes the corresponding data using a DSP core; the steps shown in labels 2 and 3 involve parallel computation, with the algorithm module deployed across 16 detection cores; as shown in labels 2 and 3. Figure 2 In each detection calculation kernel, for each partition detection grid, the processing calls the one-dimensional maximum convolution compression algorithm and the multi-grid partition parallel compression algorithm to form the basic detection data, and dynamically estimate its background and detection threshold.

[0110] The forward-looking sonar partitioned target detection device proposed in this embodiment is effective and simple, and can handle the complex underwater target detection situation caused by underwater reverberation, sound ray changes and other factors in the real marine environment. For example, in most cases of long-distance target detection, since the signal of long-distance targets is weak, this invention can also perform target detection. This is because this invention divides the detection area into grids and estimates the background and signal-to-noise ratio of the detection area corresponding to each grid to form the detection result.

[0111] This invention forms a detection image by slicing 16-bit data in the beam domain into 8-bit data slices; underwater multi-target detection in the image domain can obtain features such as the shape and shadow of underwater targets, which is more intuitive and accurate than detection in the signal domain.

[0112] This invention performs multi-target detection in the image domain. Taking into account factors such as seabed reverberation and sound propagation path, it divides a large area within the detection range of forward-looking sonar into local detection grids to form corresponding local detection images. The detection task is deployed in a multi-core concurrent thread to improve detection efficiency.

[0113] The local grid target detection algorithm of this invention statistically analyzes local noise signals and dynamically sets detection thresholds, effectively improving the detection rate when the signal of distant targets is weak.

[0114] This invention overcomes the difficulty of feature extraction in the signal domain and can detect targets by using the moment features of the target image even in the presence of strong noise signals, thereby obtaining suspected targets and improving detection efficiency.

[0115] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. An underwater target detection device based on forward-looking sonar zoning, characterized by, include: Beam domain data module (101), partitioned detection grid control module (102), compression module (103), partitioned target detection module (104), and target detection result summary module (105). The beam domain data module (101) is used to identify the parameters of the beam domain data from the signal domain processor to form forward-looking sonar beam domain data. The partitioned detection grid control module (102) divides the beam domain data into grids from two dimensions, beam and distance, according to the current detection range of the forward-looking sonar, forming a local detection grid set. The local detection grid is formed by the azimuth angle between the sonar horizontal plane and the sonar to the target and the detection distance between the sonar and the target, forming a local detection area. The compression module (103) is used to perform one-dimensional convolution on the beam domain data in each local detection grid, take the maximum value, compress and image it to form the image domain data of each local detection grid; The partitioned target detection module (104) obtains the signal-to-noise ratio in each local detection grid based on the forward-looking sonar active acoustic reflection signal and environmental noise level of the image domain data corresponding to each local detection grid to determine the detection threshold of each local detection grid, and obtains multiple underwater targets in each local detection area based on the detection threshold and the corresponding azimuth angle and detection distance. The target detection result aggregation module (105) is used to determine whether multiple underwater targets within each local detection grid have an intersection in the target location range, and merge the positions of the intersecting targets to form underwater targets; The beam domain data is the data after the sonar beam is formed, which includes target direction, sonar weight information, sonar frequency information and sonar timing information. The image domain data is obtained by compressing the beam domain data and converting it into eight-bit image data. The acquisition of the detection threshold for each local detection region includes: The beam domain data corresponding to each local detection grid is compressed and stored in the scanDetect16Buffer buffer; Obtain the mean and standard deviation (std_dev) of the beam domain data corresponding to each local detection grid within the scanDetect16Buffer. Then, the grid background for each local detection grid is: background = mean + 2 * std_dev; Based on the background grid, the detection threshold for each local detection grid is obtained, then: threshold=background / db Where: threshold is the detection threshold; db is the signal-to-noise ratio; The method of merging the positions of overlapping targets includes: Obtain the target contour area of ​​multiple underwater targets within each local detection region, and discard contours with s>400 and s<3; For a target profile that meets the conditions, calculate its centroid coordinates mX and mY, and convert the centroid coordinates into sampling point mBin and beam mBeam; The coordinate error offset range of each centroid is 10 meters. The two centroids are compared. If there is an intersection, they are merged. The merged centroid is the midpoint between the original two centroids.

2. The underwater target detection device based on forward-looking sonar partitioning according to claim 1, wherein, The current detection range of the forward-looking sonar is segmented in 100-meter ranges. The local detection grid uses 32 as an indicator, meaning that the detection grid is formed by the detection area covered by 32 beams, and each local detection grid corresponds to a 50-meter range.

3. The apparatus according to claim 1, wherein, The compression module (103) performs one-dimensional convolution on the beam domain data in each local detection grid, takes the maximum value, and then compresses the data to a size of 512*512.

4. The method according to any one of claims 1 to 3, wherein, Includes the following steps: The beam domain data module (101) receives the range and beam domain data parameters displayed on the display console and stores them in the shared buffer pool; The partition detection grid control module (102) determines the partition detection grid based on the display range and the forward sonar opening angle, and assigns the grid parameters to the corresponding partition detection task processing thread; The partition target detection module (104) extracts the corresponding beam domain data from the system data buffer and performs compressed imaging on the beam domain data; The partitioned target detection module (104) determines the detection threshold based on the signal-to-noise ratio of the detection area, performs multi-target detection in the regional image domain data, and merges similar targets; The target detection result summary module (105) summarizes the targets detected by the partition target detection module (104).

5. The method of claim 4, wherein the method further comprises: The number of processing threads for the partition detection task is set according to the full scale and the maximum opening angle of the sonar, and each thread deploys one partition detection task thread.

Images