Unmanned underwater vehicle detection method, system and equipment based on passive sonar and medium
Through the passive sonar-based unmanned submarine detection method, the unmanned submarine is identified using acoustic spectrometer and acoustic vision model, which solves the problem of unmanned submarine identification in the marine environment, and realizes efficient and accurate unmanned submarine detection and track.
Patent Information
- Application Number
- CN202411123814.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-15
- Publication Date
- 2024-11-08
AI Technical Summary
In the marine environment, existing unmanned submarine identification technology is difficult to accurately identify and track potential illegal or unknown unmanned submarines. Due to the noise and complex factors of the marine environment, the recognition accuracy rate is not high.
Using the unmanned submarine detection method based on passive sonar, the sound characteristics are extracted using the Mel frequency cepspectral coefficients to construct the sound spectrum diagram, divide and track, and combine cluster analysis and Kalman The filtering algorithm locks the suspicious target and enters a preset acoustic and visual model for identification.
It realizes efficient detection and identification of unknown underwater unmanned submarines, improves identification accuracy and adaptability, and can identify different types of unmanned submarines in different sea areas and environments, with efficient, accurate, real-time and flexible advantage.
Smart Images

Figure CN118915075A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of target passive tracking technology, and in particular to an unmanned underwater vehicle detection method, system, equipment and medium based on passive sonar. Background Art
[0002] With the continuous advancement of marine science and technology, unmanned underwater vehicles (UUVs) are increasingly used in the fields of ocean exploration, resource exploration, environmental monitoring, and military reconnaissance. UUVs have become an important tool in the field of marine science and technology due to their autonomous navigation, long-term operation, and adaptability to complex marine environments. However, with the increase in the number of UUVs, how to effectively detect potential illegal or unknown UUVs has become an urgent problem to be solved.
[0003] The marine environment is complex and changeable, including water temperature, salinity, flow rate, seabed topography and other factors, all of which may affect the search and identification of unmanned submersibles. Unmanned submersibles usually have strong concealment and complex signal characteristics, which makes it more difficult to track and identify them in the marine environment. Existing unmanned submersible identification technology mainly relies on the detection and analysis of electromagnetic signals or acoustic signals generated by unmanned submersibles, but these signals are often interfered by marine environmental noise and other factors, resulting in low recognition accuracy. Summary of the invention
[0004] In view of this, the present invention proposes an unmanned underwater vehicle detection method, system, equipment and medium based on passive sonar, which utilizes the combination of a spectrometer and passive sonar to detect unknown submersibles, and determines whether there are unknown underwater unmanned submersibles based on the obtained spectrogram, thereby realizing the discovery and tracking of the unmanned submersibles.
[0005] Based on the above purpose, the present invention proposes a method for detecting an unmanned underwater vehicle based on passive sonar, comprising: Obtain sound signals collected by passive sonar; Extracting sound features from the sound signal based on Mel-frequency cepstral coefficients to construct a spectrogram; Dividing and tracking in the spectrogram to obtain a suspicious target; The suspicious target is input into a preset acoustic-visual model to identify the suspicious target result.
[0006] In some embodiments, the step of extracting sound features from the sound signal based on Mel-frequency cepstral coefficients to construct a spectrogram comprises: Divide the sound signal into short-time frames, and construct a Mel filter bank based on a Hanning window and short-time Fourier transform; Calculate the output of the Mel filter bank, generate the corresponding filter bank coefficients and take the logarithm; Discrete cosine transform is applied to the logarithm-taking filter bank coefficients to obtain discrete cosine transform coefficients, and discrete cosine transform coefficients meeting preset conditions are selected for normalization to obtain a spectrogram.
[0007] In some embodiments, the step of dividing and tracking in the spectrogram to obtain a suspicious target includes: The data in the spectrogram is divided into suspicious mobile target groups and other static detection object groups through cluster analysis. Based on the Kalman filter algorithm, the pixel points in the spectrogram of the suspicious moving target group are calculated to lock the tracking target, and the suspicious target is obtained after comparison.
[0008] In some embodiments, the step of inputting the suspicious target into a preset acoustic-visual model to identify the suspicious target result includes: Inputting the suspicious target into a preset acoustic-visual model to perform target recognition and classification; Based on the results of target recognition and classification, it is determined whether the suspicious target result is an unknown underwater unmanned submersible.
[0009] In some embodiments, the method further comprises the following training process for the acoustic-visual model: Obtain historical sonar images for data augmentation, divide them into training sets and test sets, and preprocess the training sets and test sets; The training set is input into the convolutional neural network for training, and the classifier is used for classification prediction to obtain the training model; The test set is input into the training model for testing. When the prediction result accuracy of the test set is not lower than the preset threshold, the acoustic-visual model is obtained.
[0010] In some embodiments, the step of preprocessing the training set and the test set includes: Remove noise points in the training set and test set based on median filtering; Extract the time domain and frequency domain features of the sound signal and perform normalization processing; The training set and the test set are labeled based on the principle of sonic wave fish driving to obtain corresponding labels.
[0011] In some embodiments, the step of obtaining the sound signal collected by passive sonar includes: Select an environment that meets the requirements, turn on the passive sonar, and turn on the underwater acoustic transducer; Get the sound signal corresponding to the current moment.
[0012] The present invention proposes an unmanned underwater vehicle detection system based on passive sonar, comprising: A collection unit configured to obtain a sound signal collected by passive sonar; A construction unit configured to extract sound features from the sound signal based on Mel-frequency cepstral coefficients to construct a spectrogram; A tracking unit configured to divide and track in the sound spectrogram to obtain a suspicious target; The identification unit is configured to input the suspicious target into a preset acoustic-visual model to identify the suspicious target result.
[0013] The present invention provides a computer device, comprising: At least one processor; and a memory, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the program, the steps of the passive sonar-based unmanned underwater vehicle detection method are performed.
[0014] The present invention proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the unmanned underwater vehicle detection method based on passive sonar are executed.
[0015] The present invention has at least the following beneficial technical effects: The present invention proposes a method, system, equipment and medium for detecting an unmanned underwater vehicle based on passive sonar. The method comprises: obtaining a sound signal collected by passive sonar; extracting sound features from the sound signal based on Mel-frequency cepstral coefficients to construct a spectrogram; dividing and tracking in the spectrogram to obtain a suspicious target; inputting the suspicious target into a preset acoustic-visual model to identify the suspicious target result.
[0016] The sound signal collected by the present invention is converted into a spectrogram, and the obtained spectrogram is used to determine whether there is an unknown underwater unmanned submersible, thereby realizing the detection of the unknown submersible. Different sound wave frequencies can be selected for the working sonar image processing of the underwater acoustic transducer for identifying and detecting fish-like robots according to different operating environments. It is suitable for various types of ships and underwater robots, and has improved adaptability to different aquatic environments. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other embodiments can be obtained based on these drawings without paying creative work.
[0018] Figure 1A flow chart of the unmanned underwater vehicle detection method based on passive sonar provided by the present invention; Figure 2 A module diagram of an unmanned underwater vehicle detection system based on passive sonar provided by the present invention; Figure 3 A schematic diagram of the structure of an embodiment of a computer device provided by the present invention; Figure 4 A schematic diagram of the structure of an embodiment of a computer-readable storage medium provided by the present invention. DETAILED DESCRIPTION
[0019] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the embodiments of the present invention are further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.
[0020] It should be noted that all expressions using "first" and "second" in the embodiments of the present invention are for distinguishing two non-identical entities with the same name or non-identical parameters. It can be seen that "first" and "second" are only for the convenience of expression and should not be understood as limitations on the embodiments of the present invention. The subsequent embodiments will not explain this one by one.
[0021] The present invention proposes a method for detecting unmanned underwater vehicles based on passive sonar. Figure 1 ,include: Obtain sound signals collected by passive sonar; Extracting sound features from the sound signal based on Mel-frequency cepstral coefficients to construct a spectrogram; Dividing and tracking in the spectrogram to obtain a suspicious target; The suspicious target is input into a preset acoustic-visual model to identify the suspicious target result.
[0022] The acquisition of acoustic visual images is based on the spectrometer of passive sonar signals. A spectrometer is an instrument or system specially designed to convert sound signals into a spectrogram. A spectrogram is a spectrum representation of a sound signal, usually with time as the horizontal axis and frequency as the vertical axis. The energy or amplitude of the sound is represented by color or brightness.
[0023] The present invention uses advanced signal processing technology to achieve efficient search and identification of unmanned submersibles in the ocean. Using the preset acoustic and visual models, different types of unmanned submersibles can be accurately identified. Data can be collected and analyzed in real time to achieve real-time tracking and monitoring of unmanned submersibles. It can be expanded and upgraded according to actual needs to adapt to the search and identification tasks of different sea areas and different types of unmanned submersibles, and has the advantages of high efficiency, accuracy, real-time and flexibility.
[0024] In some embodiments, see Figure 1The step of extracting sound features from the sound signal based on the Mel-frequency cepstral coefficients and constructing a spectrogram comprises: Divide the sound signal into short-time frames, and construct a Mel filter bank based on a Hanning window and short-time Fourier transform; Calculate the output of the Mel filter bank, generate the corresponding filter bank coefficients and take the logarithm; Discrete cosine transform is applied to the logarithm-taking filter bank coefficients to obtain discrete cosine transform coefficients, and discrete cosine transform coefficients meeting preset conditions are selected for normalization to obtain a spectrogram.
[0025] Mel-frequency cepstral coefficients (MFCC), MFCC effectively reduces the dimension of features by converting spectral data into cepstral coefficients. This dimensionality reduction not only reduces the amount of calculation, but also reduces the need for storage space, thereby improving efficiency. MFCC removes some unimportant high-frequency details and noise in the original audio signal by filtering and logarithmic compression of spectral data. This denoising and filtering process enables MFCC to retain more important sound features and improve robustness and noise resistance.
[0026] Discrete Cosine Transform (DCT), DCT has a strong energy compression property, which means that it can concentrate the main energy of a signal or image on a small number of DCT coefficients. This property makes DCT particularly useful in data compression, because only a few important coefficients can be retained, while other coefficients can be ignored or approximated, thereby achieving effective data compression. DCT matches the perceptual characteristics of the human visual and auditory systems. In image processing, DCT can simulate the sensitivity of the human visual system to image brightness and frequency, thereby extracting image features that are more in line with human perceptual characteristics. In audio processing, DCT can also simulate the sensitivity of the human ear to different frequencies, thereby extracting audio features that are more in line with human auditory characteristics.
[0027] The specific process is to divide the original sound signal into short time frames, each of which lasts between 20 and 40 milliseconds to capture the time domain changes of the sound signal. Apply a Hanning window to each frame, and then perform a short-time Fourier transform to convert the time domain signal into a frequency domain signal, which will produce the spectrum information of each frame and construct a set of Mel filters that are evenly distributed on the Mel frequency scale. These filters are usually triangular filters to capture the distribution of the sound signal on the Mel frequency. Apply a Mel filter bank to the spectrum of each frame and calculate the output of each filter, which will generate a series of filter bank coefficients to represent the distribution of the sound signal on different Mel frequencies. Take the logarithm of the filter bank coefficients to simulate the human ear's perception of sound intensity. Make the features closer to the perception of the human auditory system. Apply DCT to the logarithmic filter bank coefficients to convert the frequency domain features into cepstral coefficients. Keep the first few DCT coefficients, which constitute the final MFCC feature vector. Finally, normalize the MFCC feature vector.
[0028] In some embodiments, see Figure 1 , the steps of dividing and tracking in the spectrogram to obtain suspicious targets include: The data in the spectrogram is divided into suspicious mobile target groups and other static detection object groups through cluster analysis. Based on the Kalman filter algorithm, the pixel points in the spectrogram of the suspicious moving target group are calculated to lock the tracking target, and the suspicious target is obtained after comparison.
[0029] In this embodiment, all suspicious targets in the acoustic spectrogram can be extracted, including schools of fish and unmanned underwater vehicles; Specifically, the data is pre-divided into 2 groups, and K objects are randomly selected as the initial cluster centers. Then the distance between each object and each seed cluster center is calculated, and each object is assigned to the cluster center closest to it. The cluster center and the objects assigned to them represent a cluster. Each time a sample is assigned, the cluster center of the cluster is recalculated according to the existing objects in the cluster. This process will be repeated until a certain termination condition is met. The termination condition can be that no or the minimum number of objects are reassigned to different clusters, no or the minimum number of cluster centers change again, and the error sum of squares is locally minimized. In this example, the two groups are divided into suspicious moving targets, including fish schools and underwater unmanned submersibles, and other static detection objects include terrain and bottom sediments. Subsequent tracking targets only need to be carried out for suspicious moving targets. K cluster centroid points are randomly selected to repeat the following process until convergence. For each sample i, calculate the class it should belong to. The formula is as follows: .
[0030] Among them, i is the calculated sample and j is the class.
[0031] Among them, for each class j, the centroid of the class is recalculated, and the formula is as follows: .
[0032] The Kalman filter algorithm is used to track suspicious targets. The traditional target tracking algorithm segments different areas of the image, classifies the segmented areas, and then uses the target template in adjacent images for region matching to achieve region-based tracking of the target. Since the feature point-based method is not suitable for sonar images with high noise and low resolution, the matching effect of the region-based method is poor.
[0033] Therefore, the Kalman filter method is used to achieve target tracking; by calculating the pixel points in the image, the suspicious targets of all tracked targets are locked. First, the image before the underwater acoustic transducer works is recorded to detect the number of pixels of all suspicious targets and record them in the array aim[]. After the underwater acoustic transducer works, the continuous acoustic image is scanned frame by frame, all targets are tracked and the number of their pixels is calculated and compared with the value in the aim[] array. When the number of pixels is reduced by 40% and maintained for a certain period of time, it will be identified as a school of fish and will no longer be tracked. The target that is still being tracked after the detection time t will be marked as a suspicious target.
[0034] Kalman filtering can effectively handle noise and uncertainty, and can provide accurate estimates even when there is noise in the observations. It can process observations in real time and provide the latest estimates. In target tracking, this means that the algorithm can quickly respond to changes in the position and state of the target, thereby achieving real-time tracking effects. Compared with other complex filtering algorithms, Kalman filtering has a relatively small amount of computation. This enables it to maintain high real-time performance when processing a large number of observations, and is particularly suitable for tracking scenarios that require fast responses.
[0035] Based on the values of each measurement at different times, the joint distribution at each time can be considered to generate an estimate of the unknown variable. This method is more accurate than the estimation method based on only a single measurement, and therefore can provide higher-precision tracking results.
[0036] In some embodiments, see Figure 1 The step of inputting the suspicious target into a preset acoustic-visual model and identifying the suspicious target result comprises: Inputting the suspicious target into a preset acoustic-visual model to perform target recognition and classification; Based on the results of target recognition and classification, it is determined whether the suspicious target result is an unknown underwater unmanned submersible.
[0037] Track the audio of the suspicious target; judge the tracked audio and extract the suspicious target; finally, the detection machine equipped with a high-definition camera uses the convolutional neural network CNN to identify the target and further determine whether the suspicious target is an unknown underwater unmanned submersible.
[0038] Once the target is determined to be an unmanned underwater vehicle, the location information of the unmanned underwater vehicle will be accurately calculated, and the location information, time and other related information of the discovered unmanned underwater vehicle will be notified to relevant personnel in real time through various means such as satellite communication, radio communication, and network communication.
[0039] In some embodiments, see Figure 1 The method also includes the following training process for the preset acoustic-visual model: Obtain historical sonar images for data augmentation, divide them into training sets and test sets, and preprocess the training sets and test sets; The training set is input into the convolutional neural network for training, and the classifier is used for classification prediction to obtain the training model; The test set is input into the training model for testing. When the prediction result accuracy of the test set is not lower than the preset threshold, the acoustic-visual model is obtained.
[0040] The convolutional neural network model chooses to use pytorch to build a convolutional neural network. From bottom to top, the first input image is image, which passes through a convolution layer, and then the convolution information is processed by pooling. The maxpooling method is used here. After the same processing, the second processed information is passed to two fully connected neural layers, which is also a general two-layer neural network layer. Finally, a classifier is connected for classification prediction.
[0041] Through convolution and pooling operations, local features of the image can be extracted with good position invariance. This means that even if the target in the image is displaced, rotated or scaled to a certain extent, CNN can still accurately identify the target. Due to the sharing of convolution kernels and the downsampling operation of pooling, CNN can greatly reduce the number of parameters of the model. This not only reduces the complexity and computational cost of the model, but also helps prevent overfitting and improve the generalization ability of the model.
[0042] In some embodiments, see Figure 1 , the step of preprocessing the training set and the test set includes: Remove noise points in the training set and test set based on median filtering; Extract the time domain and frequency domain features of the sound signal and perform normalization processing; The training set and the test set are annotated based on the principle of sonic wave fish driving to obtain corresponding labels.
[0043] Based on the processing of the acoustic vision system, the original sonar images are used for data augmentation, and the sample sets required for training and testing the models are obtained. The noise points are removed by median filtering. The time domain and frequency domain characteristics of the sound signal are extracted using the Mel-Frequency Cepstral Coefficients (MFCC). The fish schools and fish-like robots in the sample set are manually labeled and divided according to the different reactions of the fish schools and unmanned submersibles caused by the principle of sound wave driving fish, and the label maps required for testing and training the models are obtained.
[0044] The sound data is preprocessed, including noise reduction and data enhancement (such as random speed change or noise addition) in addition to normalization, to improve the robustness of the model.
[0045] Normalized data has similar scales, which helps most machine learning algorithms converge faster. Because the algorithm needs to calculate the gradient of the loss function during the optimization process, normalization ensures that the gradient is valid in each iteration, thereby speeding up training. Noise reduction can help the model better cope with noise interference in the actual environment. Models trained with denoising usually perform better on noisy test data, thereby enhancing the robustness of the model. It helps to extract useful information more accurately and improve data quality. Data enhancement technology can increase the amount of data by transforming and expanding the original data. Through data enhancement, the model can be exposed to more data changes and diversity. Similar to denoising, data enhancement can also reduce the risk of overfitting. It is difficult to collect enough real data. Through data enhancement technology, the data set can be expanded without adding additional costs, thereby saving data collection costs.
[0046] In some embodiments, see Figure 1 , the step of obtaining the sound signal collected by passive sonar includes: Select an environment that meets the requirements, turn on the passive sonar, and also turn on the hydroacoustic transducer; Get the sound signal corresponding to the current moment.
[0047] Passive sonar sensors require the surrounding environment to be as pure as possible, reduce miscellaneous sounds, achieve more accurate sound signal recording, and eliminate early noise points. Through the spectrometer and passive sonar mounted on the underwater robot, the sound source level of the hydroacoustic transducer is greater than 140db, the frequency is 400-1000Hz, and the frequency is adjustable. The hydroacoustic transducer uses the characteristics of sound waves propagating in water to achieve the positioning and navigation of underwater targets. By measuring the propagation time and direction of the sound waves, the position of the target can be determined. This is a key technology for achieving precise navigation and positioning for underwater navigation, underwater robots, submersibles and other equipment. The hydroacoustic transducer will disperse the fish in a direction without driving away the bionic unmanned submersible that may be hidden in the fish. The sound signal collected by the passive sonar is further converted into a spectrogram, waveform graph or frequency graph with sound characteristics.
[0048] According to another aspect of the present invention, a passive sonar-based unmanned underwater vehicle detection system is proposed. Figure 2 ,include: The acquisition unit 100 is configured to obtain a sound signal collected by passive sonar; A construction unit 200 is configured to extract sound features from the sound signal based on Mel-frequency cepstral coefficients to construct a spectrogram; A tracking unit 300, configured to divide and track in the spectrogram to obtain a suspicious target; The identification unit 400 is configured to input the suspicious target into a preset acoustic-visual model to identify the suspicious target result.
[0049] The sound signal collected by the present invention is converted into a spectrogram, and the obtained spectrogram is used to determine whether there is an unknown underwater unmanned submersible, thereby realizing the detection of the unknown submersible. Different sound wave frequencies can be selected for the working sonar image processing of the underwater acoustic transducer for identifying and detecting fish-like robots according to different operating environments. It is suitable for various types of ships and underwater robots, and has improved adaptability to different aquatic environments.
[0050] Based on the same inventive concept, according to another aspect of the present invention, Figure 3 As shown, an embodiment of the present invention further provides a computer device 30, which includes a processor 310 and a memory 320. The memory 320 stores a computer program 321 that can be run on the processor. When the processor 310 executes the program, the steps of the above method are performed.
[0051] Based on the same inventive concept, according to another aspect of the present invention, Figure 4 As shown, an embodiment of the present invention further provides a computer-readable storage medium 40, which stores a computer program 410 for executing the above method when executed by a processor.
[0052] The embodiment of the present invention may also include a corresponding computer device. The computer device includes a memory, at least one processor, and a computer program stored in the memory and executable on the processor, and the processor executes any one of the above methods when executing the program.
[0053] The memory, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer executable programs and modules, such as program instructions / modules in the embodiments of the present application. The processor executes various functional applications and data processing of the device by running the non-volatile software programs, instructions and modules stored in the memory, that is, implementing the above method.
[0054] The memory may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the device, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In an embodiment, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the local module via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0055] Finally, it should be noted that a person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the storage medium of the program can be a disk, an optical disk, a read-only storage memory (ROM) or a random access memory (RAM), etc. The above-mentioned computer program embodiments can achieve the same or similar effects as the corresponding above-mentioned arbitrary method embodiments.
[0056] It will also be appreciated by those skilled in the art that various exemplary logic blocks, modules, circuits and algorithm steps described in conjunction with the disclosure herein can be implemented as electronic hardware, computer software or a combination of the two. In order to clearly illustrate this interchangeability of hardware and software, a general description has been given to the functions of various schematic components, blocks, modules, circuits and steps. Whether this function is implemented as software or hardware depends on specific applications and the design constraints imposed on the entire system. Those skilled in the art can implement the function in various ways for each specific application, but this implementation decision should not be interpreted as causing a departure from the disclosed scope of the embodiments of the present invention.
[0057] The above are exemplary embodiments disclosed in the present invention, but it should be noted that various changes and modifications may be made without departing from the scope of the embodiments disclosed in the present invention as defined in the claims. The functions, steps and / or actions of the method claims according to the disclosed embodiments described herein do not need to be performed in any particular order. The serial numbers of the embodiments disclosed in the above embodiments of the present invention are for description only and do not represent the advantages and disadvantages of the embodiments. In addition, although the elements disclosed in the embodiments of the present invention may be described or required in individual form, they may also be understood as multiple unless explicitly limited to the singular.
[0058] It should be understood that, as used herein, the singular forms "a", "an" are intended to include the plural forms as well, unless the context clearly supports an exception. It should also be understood that, as used herein, "and / or" refers to any and all possible combinations including one or more of the associated listed items.
[0059] A person skilled in the art should understand that the discussion of any of the above embodiments is only exemplary and is not intended to imply that the scope of the disclosure of the embodiments of the present invention (including the claims) is limited to these examples; under the concept of the embodiments of the present invention, the technical features in the above embodiments or different embodiments can also be combined, and there are many other changes in different aspects of the above embodiments of the present invention, which are not provided in detail for the sake of simplicity. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present invention should be included in the protection scope of the embodiments of the present invention.
Claims
1. An unmanned underwater vehicle detection method based on passive sonar, characterized in that: include: Obtain sound signals collected by passive sonar; Extracting sound features from the sound signal based on Mel-frequency cepstral coefficients to construct a spectrogram; Dividing and tracking in the acoustic spectrogram to obtain a suspicious target; The suspicious target is input into a preset acoustic-visual model to identify the suspicious target result.
2. The passive sonar-based unmanned underwater vehicle detection method according to claim 1, characterized in that: The step of extracting sound features from the sound signal based on the Mel-frequency cepstral coefficients to construct a spectrogram comprises: Divide the sound signal into short-time frames, and construct a Mel filter bank based on a Hanning window and short-time Fourier transform; Calculate the output of the Mel filter bank, generate the corresponding filter bank coefficients and take the logarithm; Discrete cosine transform is applied to the logarithm-taking filter bank coefficients to obtain discrete cosine transform coefficients, and discrete cosine transform coefficients meeting preset conditions are selected for normalization to obtain a spectrogram.
3. The passive sonar-based unmanned underwater vehicle detection method according to claim 1, characterized in that: The steps of dividing and tracking the spectrogram to obtain a suspicious target include: The data in the spectrogram is divided into suspicious mobile target groups and other static detection object groups through cluster analysis. Based on the Kalman filter algorithm, the pixel points in the spectrogram of the suspicious moving target group are calculated to lock the tracking target, and the suspicious target is obtained after comparison.
4. The method for detecting an unmanned underwater vehicle based on passive sonar according to claim 1, characterized in that: The step of inputting the suspicious target into a preset acoustic-visual model to identify the suspicious target result comprises: Inputting the suspicious target into a preset acoustic-visual model to perform target recognition and classification; Based on the results of target recognition and classification, it is determined whether the suspicious target result is an unknown underwater unmanned submersible.
5. The method for detecting an unmanned underwater vehicle based on passive sonar according to claim 1, characterized in that: The method also includes the following training process for the acoustic-visual model: Obtain historical sonar images for data augmentation, divide them into training sets and test sets, and preprocess the training sets and test sets; The training set is input into the convolutional neural network for training, and the classifier is used for classification prediction to obtain the training model; The test set is input into the training model for testing, and when the prediction result accuracy of the test set is not lower than a preset threshold, the acoustic-visual model is obtained.
6. The passive sonar-based unmanned underwater vehicle detection method according to claim 5, characterized in that: The step of preprocessing the training set and the test set comprises: Remove noise points in the training set and test set based on median filtering; Extract the time domain and frequency domain features of the sound signal and perform normalization processing; The training set and the test set are labeled based on the principle of sonic wave fish driving to obtain corresponding labels.
7. The method for detecting an unmanned underwater vehicle based on passive sonar according to claim 1, characterized in that: The step of obtaining the sound signal collected by passive sonar comprises: Select an environment that meets the requirements, turn on the passive sonar, and turn on the underwater acoustic transducer; Get the sound signal corresponding to the current moment.
8. The unmanned underwater vehicle detection system based on passive sonar is characterized by: include: A collection unit configured to obtain a sound signal collected by passive sonar; A construction unit configured to extract sound features from the sound signal based on Mel-frequency cepstral coefficients to construct a spectrogram; A tracking unit configured to divide and track in the sound spectrogram to obtain a suspicious target; The identification unit is configured to input the suspicious target into a preset acoustic-visual model to identify the suspicious target result.
9. A computer device comprising: at least one processor; and a memory storing a computer program executable on the processor, wherein the processor executes the steps of the passive sonar-based unmanned underwater vehicle detection method as described in any one of claims 1 to 7 when executing the program.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the passive sonar-based unmanned underwater vehicle detection method as described in any one of claims 1 to 7 are performed.
Citation Information
Cited By
Underwater sound signal positioning method and device and storage medium
CN120370321A
