Underwater target identification module based on simulink

Abstract

The invention provides an underwater target identification module based on simulink, which can solve the problems that an underwater target identification device in the current military product and civil product industry is long in trial-manufacture period, large in actual flight verification resource consumption and the like. The digital twinning module serves as a digital twinning module of an underwater target identification module physical model machine, the digital twinning module comprises a signal conditioning sub-module and a data processing module, the signal conditioning sub-module corresponds to an analog circuit part in the physical model machine, and the signal processing sub-module corresponds to a microprocessor unit part in the physical model machine; the signal conditioning sub-module at the front end is used for carrying out narrowband filtering and detection processing on the ship sound signal; and the data processing module is used for receiving the ship magnetic field signal, the water pressure signal and the conditioned sound field signal, and finally outputting a target identification result in real time after processing.

Description

Technical Field

[0001] The invention relates to signal detection technology, underwater acoustic signal processing and ship physical field analysis technology, and belongs to the field of underwater acoustic detection technology and ship physical field identification. Background Art

[0002] Underwater target recognition technology is widely used in domestic military and civilian industries, among which the recognition technology involving ship physical field signals is more studied and used by the underwater weapons industry. The traditional product development process is "drawing + processing + testing", which has a long cycle and consumes a lot of human resources and funds. It is often necessary to wait until the physical prototype is completed before field actual navigation tests can be carried out to verify the performance indicators of the product. Once major technical design problems are exposed in field tests, it is difficult to remedy them in time, and "delaying progress, reducing indicators, and increasing expenses" become inevitable problems. Summary of the invention

[0003] In view of this, the present invention proposes an underwater target recognition module based on Simulink, which can solve the problems of long trial production cycle of underwater target recognition devices and high consumption of actual navigation verification resources in the current military and civilian industries.

[0004] The specific technical solutions are as follows:

[0005] A simulink-based underwater target recognition module, as a digital twin module of a physical prototype of the underwater target recognition module, comprises a signal conditioning submodule and a data processing module, the signal conditioning submodule corresponds to the analog circuit part in the physical prototype, and the signal processing submodule corresponds to the microprocessor unit part in the physical prototype; the front-end signal conditioning submodule is used for performing narrow-band filtering and detection processing on ship acoustic signals; the data processing module is used for receiving ship magnetic field signals, water pressure signals, and conditioned sound field signals, and finally outputs target recognition results in real time after processing.

[0006] Furthermore, the attenuation of the acoustic signal is determined by the propagation distance, and a step signal is used to simulate the depth signal input, which is forced to be converted into Uint16 type and down-sampled to simulate the AD sampling process, and then input into the data processing module.

[0007] Furthermore, in order to control the operation timing, a counter link is introduced to provide a synchronous clock to the data processing module so that it can output the judgment result in every fixed period.

[0008] Furthermore, the data processing module is used for sampling of acoustic, magnetic, water pressure signals and depth information, acoustic signal gain control, acoustic signal processing, magnetic signal processing, water pressure signal processing, and finally outputting the recognition results.

[0009] Further, for the acoustic signal gain control, it refers to the real-time gain adjustment according to the acquired depth information and the amplitude of the acoustic signal, which is used to eliminate the saturation phenomenon of the acoustic field signal affected by the deployment depth. The acoustic signal processing includes three parts, namely slope judgment, threshold judgment, and frequency-domain energy judgment. Among them, the slope judgment is used to identify whether there is a target. Specifically, the rising trend of the acquired acoustic signal is used as one of the identification bases. When the target approaches the deployment position, the energy of the full frequency band increases. Therefore, an energy threshold is set to be greater than the underwater normal noise energy and less than the energy peak generated by the passing of a ship; frequency-domain energy judgment. The radiated noise of a ship includes engine noise and propeller noise. Both types of noise exist in a specific frequency range and are concentrated in the low-frequency band. Therefore, after filtering the acoustic field signal, by comparing the energy magnitudes of these frequency bands, it can be determined whether this group of signals is a ship signal;

[0010] The magnetic signal processing includes three parts, namely magnetic declination angle judgment, magnetic induction intensity threshold judgment, and magnetic field component correlation judgment; the magnetic declination angle judgment means that by using the change trend of the magnetic declination angle caused when the target passes, it is further judged whether there is a target; the magnetic induction intensity is relatively low when there is no target, which is consistent with the acoustic field signal. By setting a threshold value greater than the magnetic background signal and less than the magnetic induction intensity of the target as the discrimination basis for the target; the magnetic field component correlation judgment means that since the magnetic induction intensity of the target is strongly correlated with its own magnetic field distribution, the three components of the actually acquired magnetic induction intensity of the target do not have strong correlation, that is, they will not change in phase simultaneously. By calculating the correlation coefficients of the three components, the target can be distinguished from the interference signal;

[0011] The water pressure signal processing includes two parts, water pressure slope change detection and water pressure threshold detection; when the target passes, the change slope of the water pressure signal here is greater than that when there is no target, and there will be a water pressure value higher than the normal state. By setting the slope threshold and the water pressure value threshold according to the characteristics of the passing of the target, they can be used as the criteria for target recognition;

[0012] All three signals can be used as the basis for target recognition.

[0013] Further, the data processing module can also perform target recognition through fusion decision-making, that is, by means of weight distribution for the three signals, the target recognition result is finally obtained.

[0014] Beneficial effects

[0015] 1. The underwater target recognition module designed based on Simulink described in the present invention has been verified by virtual tests on multi-physical field signals generated during the actual navigation of a certain type of real ship, and can effectively recognize actual ship targets.

[0016] 2. The present invention is a digital prototype of an underwater target recognition module designed based on Simulink. It has the ability to recognize underwater targets and excellent digital characteristics, and can modify key working parameters in the prototype in real time. By injecting multi-physical field data of ships under different working conditions, virtual experimental verification and optimization of the effectiveness of signal processing algorithms are carried out, and the optimization results are timely fed back to designers, which can greatly improve the development efficiency of products.

[0017] 3. The present invention uses the method of constructing a simulation model with Simulink tools and integrating a digital system prototype to carry out simulation design on the target recognition module in the synchronous development model project. At the same time, based on the historical in-field actual navigation data of underwater targets, virtual experiments are used instead of actual navigation experiments to realize the iteration of key product parameters and the verification of core indicators, provide guidance for the development of subsequent physical prototypes, greatly improve the design and verification efficiency of model projects, and significantly shorten the development cycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 Original signal of the sound field of a certain type of ship

[0019] Figure 2 Original signal of the magnetic field of a certain type of ship

[0020] Figure 3 Original signal of the hydrostatic pressure field of a certain type of ship

[0021] Figure 4 Digital prototype model

[0022] Figure 5 Sampling processing and recognition results of the sound field signal of a certain type of ship

[0023] Figure 6 Sampling and recognition results of the magnetic field signal of a certain type of ship

[0024] Figure 7 Sampling and recognition results of the hydrostatic pressure field signal of a certain type of ship DETAILED DESCRIPTION OF THE INVENTION

[0025] An underwater target recognition module designed based on Simulink is invented. Ship physical field signals such as sound, magnetism, and hydrostatic pressure are loaded into the digital prototype system. After narrowband filtering and detection preprocessing of the signals by the front-end signal conditioning sub-module, the three signals are simultaneously input into the back-end signal processing sub-module. The signal processing sub-module performs intelligent fusion processing on the three physical field signals and finally outputs the target recognition result in real time.

[0026] The usage method of this digital prototype includes the following steps:

[0027] 1. Process the original multi - physical - field signal. To make the injected acoustic field, magnetic field, and water pressure field signals meet the requirements of different sampling rates and be synchronously loaded to the input end of the digital prototype, it is necessary to add time - series information to the original signal sequence. This process is implemented through the timeseries function built in MATLAB. To obtain more intuitive test results, here, 2 minutes before and 2 minutes after the ship passes the product layout point, that is, a total of 4 minutes of measured multi - physical - field signal data are selected for verification. Here, the synchronous signals of the acoustic field, magnetic field, and water pressure field obtained from a certain type of surface ship during sea trials are selected for research. The original signals of the three physical fields collected in the experiment are as shown in the appendix Figures 1-3 as follows. The magnetic field signal contains six components. The first three channels are the X - axis, Y - axis, and Z - axis components of the static induction magnetic field, and the last three channels are the pitch angle, roll angle, and yaw angle of the prototype in the underwater deployment state.

[0028] 2. The digital prototype designed by the present invention is as shown in the appendix Figure 4 as follows. It is mainly composed of a signal conditioning sub - module at the front end and a signal processing sub - module at the rear end. The signal conditioning sub - module corresponds to the analog circuit part in the physical prototype, and the signal processing sub - module corresponds to the micro - processor unit part in the physical prototype. Since the data characteristics of different physical fields are different, the injection methods of the three types of data are also different. First, for the injection operation of the acoustic field signal, ① is the original sampled acoustic signal with the time series introduced in step 1. Since this acoustic signal has been amplified when passing through the collector, here, to restore the original signal, it is necessary to attenuate it first. To improve the system simulation operation speed, a single - type forced conversion link is introduced here after attenuation to reduce the operation amount of the whole system. The sampling frequency of the collector is much greater than the sampling frequency required for the signal processing link, so a zero - order hold link is introduced here. To reduce the sampling frequency of the original signal and further improve the simulation operation speed. For the acoustic field signal, in addition to containing time - domain information, it also has rich frequency - domain information, but the frequency - domain information required for target recognition generally only exists in some specific frequency bands. Therefore, in the signal conditioning module, the components outside the required frequency bands need to be removed, that is, narrow - band filtering processing is performed in the signal conditioning module. In this design, the original acoustic field data is decomposed into eight narrow - band signals of different frequency bands. And to facilitate the signal processing module to analyze, a peak detection module is introduced after the filtering module. The narrow - band filtering module and the detection module used here are both implemented through the code edited by S - function.

[0029] Since digital filtering will cause large signal jumps and ramps, a smoothing processing link ② is introduced here by imitating the physical product. Its processing method is as shown in formula (1):

[0030] M = M1 * 0.98+M2 * 0.02 (1)

[0031] In the formula, M is the output value after the smoothing process of the sound field signal, M1 is the sampling value of the previous cycle, and M2 is the sampling value of the current cycle. The processed signal will have a certain delay, but the delay is small and can be ignored. The processing result of the sound field signal obtained here is an analog quantity, which is the AD sampling process of the analog processing module. Here, a Uint16 forced conversion link is introduced at the back end. The sound field signal obtained after the above processing can be injected into the signal processing module for parsing.

[0032] The injection of the magnetic field signal ③ is simpler than that of the sound field. Since the magnetic field data collected by the collector is serial port data, that is, digital signal, no extra processing is required. Consistent with the sound field signal, here to improve the simulation operation speed, the magnetic field signal is converted into single format and a zero-order hold downsampling operation is performed.

[0033] The acquisition process of the hydrostatic pressure field signal ④ is similar to that of the sound field signal. However, since its maximum period is less than 1 Hz, there is almost no effective frequency domain information, and only time domain analysis is required. In the actual processing process, the gain of the hydrostatic pressure signal remains unchanged. Here, it is directly amplified, and then downsampling operation is performed, and a Uint16 forced conversion link is introduced for the analog AD sampling process.

[0034] It can be seen from the sonar equation that the propagation loss TL of the spherical wave is determined by Equation (2):

[0035] TL = 20lg(r) + αr (2)

[0036] In the formula, r is the propagation distance of the spherical wave, and α is the seawater sound absorption coefficient, which can be ignored for low-frequency signals. It can be seen from Equation (2) that the attenuation of the sound signal is mainly determined by the propagation distance. Therefore, the deployment depth will directly affect the sound pressure level acting on the surface of the receiver. To extract the effective components of the sound field signal from the noise, it needs to be amplified. And to prevent signal saturation, the gain of the signal needs to be dynamically adjusted according to the deployment depth. Therefore, the deployment depth of the product will directly affect the signal processing result. In the figure, ⑤ uses a step signal to simulate the input of the depth signal. Since the output value of the depth sensor is generally a voltage value, that is, an analog signal, here it is also forced to be converted into Uint16 type for the analog AD sampling process, as well as downsampling processing.

[0037] To control the operation timing of the processor, a counter link ⑥ is introduced here. The purpose of this link is to provide a synchronous clock for the data processing sub-module so that it can output a discrimination result in each fixed cycle.

[0038] 3. The data processing module ⑦ is composed of an S-function, which corresponds to the microprocessor unit in the physical prototype. Its main tasks are as follows: sampling of three-field signals and depth information, sound field signal gain control, acoustic signal processing, magnetic signal processing, hydrostatic pressure signal processing, and fusion decision-making algorithm.

[0039] The front-end signal conditioning module has completed the preprocessing of the signals. In the data processing module, only the sampling operation of the required signals needs to be completed and stored in the corresponding arrays. Among them, the sound field signal is affected by the deployment depth and shows a saturation phenomenon. Therefore, it is necessary to perform real-time gain adjustment according to the acquired depth information and the amplitude of the acoustic signal to facilitate the subsequent algorithm process.

[0040] The acoustic signal processing algorithm consists of three parts, namely slope judgment, threshold judgment, and frequency-domain energy judgment. As Figure 1 shown, the target passing sound field signal of a certain type of ship is given here. It can be found that when there is no target, the sound field signal has no obvious fluctuation, that is, the time periods from 0 to 50 s and after 200 s in the figure. When the target approaches from far to near and then moves away, the envelope of the sound field signal shows a relatively obvious change, that is, the time period from 50 to 200 s shown. Therefore, the rising trend of the acquired signal can be used as one of the identification bases. Secondly, when the target approaches the deployment position, the energy of the full frequency band has a relatively obvious increase. Therefore, an energy threshold is set here so that it exceeds the underwater normal noise energy and is less than the energy peak generated by the passing of the ship, which can also be used as a discrimination basis. However, due to the complex marine environment, when the sea conditions are relatively severe, such as when a typhoon passes by, the above phenomena can also occur, which will affect the target discrimination result. To solve this problem, a third criterion, namely frequency-domain energy judgment, needs to be introduced. The radiated noise of the ship is mainly caused by the ship's power, that is, engine noise and propeller noise. The characteristics of these two types of noise are that they both exist in a specific frequency range and are concentrated in the low-frequency band. Therefore, after filtering the sound field signal, comparing the energy magnitudes of these frequency bands can determine whether this group of signals is a ship signal, and it can effectively distinguish it from interference signals.

[0041] The magnetic signal processing algorithm consists of three parts, namely magnetic declination angle judgment, magnetic induction intensity threshold judgment, and magnetic field component correlation judgment. As Figure 2As shown, the magnetic field passing characteristics of the same type of ship are given here. It can be seen that when there is no target passing, the three magnetic field components are almost 0 and remain unchanged. When a target passes, the three components will have obvious fluctuations. For the magnetic declination angle judgment, that is, the target azimuth judgment, before judging the target azimuth, it is first necessary to determine the attitude of the prototype itself. Therefore, the pitch angle, roll angle, and yaw angle of the prototype itself are collected here. By calculating the state transition matrix, the prototype coordinate system can be transformed into the geodetic coordinate system, and then the target azimuth can be determined through the changes in the three magnetic field components. For the true target passing characteristics, the angle will show a process of gradually shrinking to a certain value and then continuously increasing until it is far away. Moreover, the change rate of this angle is related to the target navigation speed and will not change suddenly. Therefore, the change trend of the magnetic declination angle can be used as one of the criteria. Secondly, the magnetic induction intensity is relatively low when there is no target, which is consistent with the acoustic field signal. A threshold value greater than the magnetic background signal and less than the target can be set as the discrimination basis for the target. Finally, since the magnetic induction intensity of the target is strongly correlated with its own magnetic field distribution, the three components of the magnetic induction intensity of the target actually collected do not have strong correlation, that is, the phase changes will not occur simultaneously. The target can be distinguished from the interference signal by calculating the correlation coefficient of the three components.

[0042] The hydrostatic pressure signal algorithm consists of two parts: hydrostatic pressure slope change detection and hydrostatic pressure threshold detection. As Figure 3 shown, the hydrostatic pressure passing characteristics of the same type of ship are given here. When there is no target, it shows a relatively regular change trend. When a target passes, a large mutation will occur. Here, the change slope of the hydrostatic pressure signal is greater than that when there is no target, and there will be a hydrostatic pressure value higher than the normal state. Reasonable slope threshold and hydrostatic pressure value threshold can be set according to the target passing characteristics as the criteria for target recognition.

[0043] The above three physical fields can all be used as the basis for target recognition, but each has its limitations. To organically combine the information of these three physical fields, a fusion processing algorithm is still needed. This fusion processing algorithm includes most of the characteristics of target passing and distinguishes the target from artificial interference. For example, when a typhoon passes by, it will generate large acoustic field and hydrostatic pressure field interference, but has little impact on the magnetic field; when there is artificial interference signal generated by a small target, it will have a large interference on the acoustic field and magnetic field, but will not cause obvious hydrostatic pressure changes; when a real ship passes by, it will simultaneously cause changes in the acoustic, magnetic, and hydrostatic pressure fields and meet the requirements of the above algorithm criteria. Reasonable weight distribution is carried out on the three physical field signals, and comprehensive analysis is carried out to finally obtain the target recognition result. Since the form of pure digital prototype is adopted in this design, it is relatively convenient and easy to change and verify each weight and threshold, greatly improving the simulation and parameter iteration speed and the development efficiency.

[0044] After injecting the multi-physical field data processed in step 2, the target recognition result of this digital prototype is as attachedFigures 5-7 As shown, the moment when the target recognition result curve jumps is the recognition moment. It can be seen that before the surface ship reaches the beam position of the target recognition device, the digital prototype can accurately recognize and judge it as a valid target, and the design result is satisfactory.

Claims

1. A simulink-based underwater target recognition module, as a digital twin module of a physical prototype of the underwater target recognition module, characterized in that: It includes a signal conditioning submodule and a data processing module. The signal conditioning submodule corresponds to the analog circuit part in the physical prototype, and the signal processing submodule corresponds to the microprocessor unit part in the physical prototype. The front-end signal conditioning submodule is used to perform narrow-band filtering and detection processing on the ship sound signal. The data processing module is used to receive the ship's magnetic field signal, water pressure signal, and conditioned sound field signal, and finally output the target recognition result in real time after processing; the signal conditioning submodule and the data processing module are implemented by S-function.

2. The underwater target recognition module based on Simulink according to claim 1, characterized in that: The attenuation of the acoustic signal is determined by the propagation distance. A step signal is used to simulate the depth signal input, which is forced to be converted into Uint16 type and downsampled to simulate the AD sampling process, and then input into the data processing module.

3. The underwater target recognition module based on Simulink according to claim 1, characterized in that: In order to control the operation timing, a counter link is introduced to provide a synchronous clock to the data processing module so that it can output the judgment result in every fixed period.

4. The underwater target recognition module based on Simulink according to claim 1, characterized in that: The data processing module is used for sampling of acoustic, magnetic, water pressure signals and depth information, acoustic signal gain control, acoustic signal processing, magnetic signal processing, water pressure signal processing, and finally outputting recognition results.

5. The underwater target recognition module based on Simulink according to claim 1, characterized in that: Acoustic signal gain control: refers to the real-time gain adjustment based on the collected depth information and acoustic signal amplitude, which is used to eliminate the saturation phenomenon of the sound field signal caused by the deployment depth; Acoustic signal processing includes three parts, namely slope judgment, threshold judgment, and frequency domain energy judgment, among which the slope judgment is used to identify whether there is a target, and specifically the rising trend of the collected acoustic signal is used as one of the identification bases; When the target approaches the deployment position, the energy of the entire frequency band rises and changes, so the energy threshold is set to exceed the normal underwater noise energy and be less than the energy peak generated by the passing ship; Frequency domain energy judgment. Ship radiated noise includes engine noise and propeller noise. Both types of noise exist in a specific frequency range and are concentrated in the low frequency band. Therefore, after filtering the sound field signal, the energy of these frequency bands is compared to determine whether the group of signals is a ship signal; Magnetic signal processing includes three parts, namely, magnetic deflection angle judgment, magnetic induction intensity threshold judgment, and magnetic field component correlation judgment; magnetic deflection angle judgment refers to using the change trend of the magnetic deflection angle caused by the target passing by to judge whether there is a target; the magnetic induction intensity when there is no target is low, which is consistent with the sound field signal. By setting a threshold value greater than the magnetic background signal and less than the target magnetic induction intensity, it is used as the basis for target discrimination; magnetic field component correlation judgment refers to that since the target's magnetic induction intensity is strongly correlated with its own magnetic field distribution, the three components of the target magnetic induction intensity actually collected do not have a strong correlation, that is, the phase will not change at the same time. The target is distinguished from the interference signal by calculating the correlation coefficient of the three components; Water pressure signal processing includes two parts: water pressure slope change detection and water pressure threshold detection. When a target passes by, the change slope of the water pressure signal here is greater than when there is no target, and there will be a water pressure value higher than the normal value. The slope threshold and water pressure value threshold can be set according to the target passing characteristics as the criterion for target recognition. All three signals can be used as a basis for target identification.

6. The underwater target recognition module based on Simulink according to any one of claims 1 to 5, characterized in that: The data processing module can also perform target recognition through fusion judgment, that is, to obtain the target recognition result by allocating weights to the three signals.

Images