Active and passive combined tracking method and system based on volume Kalman filtering
By fusing data from active and passive sonar using volumetric Kalman filtering, the limitations of each method in underwater target tracking are overcome, achieving high-precision and stable underwater target tracking. This method is applicable to fields such as ocean exploration, ship defense, submarine surveillance, and underwater communication.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-15
- Publication Date
- 2026-04-03
AI Technical Summary
Existing active and passive sonar each have their limitations in underwater target tracking. Active sonar has poor concealment and low detection accuracy at long distances, while passive sonar has insufficient observation information and is easily affected by environmental noise, making it difficult to achieve stable, continuous and high-precision tracking.
A joint active and passive tracking method based on capacitive Kalman filtering is adopted. By constructing a target state model and their respective observation models, the acoustic wave data of active sonar and passive sonar are processed separately. An information fusion strategy is designed using variance fusion rules to fuse the estimated data of the two to obtain the fused target state vector estimation result.
It improves the accuracy of target state estimation, optimizes the observation dimension, sampling frequency and acoustic data transmission loss, achieves high-precision tracking performance in different distance ranges, reduces communication bandwidth burden and improves response efficiency.
Smart Images
Figure CN121784746A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of underwater target tracking technology, and in particular to a combined active and passive tracking method and system based on capacitive Kalman filtering. Background Technology
[0002] Underwater target tracking is a key technology for dynamically estimating the motion state of a target using time-series observation information. It is widely used in fields such as ocean exploration, ship defense, submarine surveillance, underwater communication, and intelligent equipment navigation. Its core idea is to continuously observe the target using a sensor array deployed on an underwater platform, extract the target's characteristic parameters (such as distance, direction, and velocity), and combine the target's motion model with the sensor observation model, using filtering or optimization algorithms to recursively estimate the target's motion state in space.
[0003] In practical applications, sonar systems are commonly used for underwater target observation. Based on the different operating methods of the sensors, sonar can be divided into two types: active sonar and passive sonar. Active sonar is an acoustic system that detects targets by actively emitting sound wave signals and receiving the echoes. Its basic principle is to emit a preset type of pulse signal into the sea area within a certain time period, and by receiving the echo signals reflected from the target, extract the target's distance and direction estimation information, thereby constructing the target's observation status. This type of system can provide high-resolution spatial positioning information, and is particularly suitable for high-precision target tracking tasks at medium to short ranges. Unlike active sonar, passive sonar does not actively emit signals, but passively receives noise signals radiated by the target or reflected signals from external sound sources at the target location. By processing the received signals, the target's direction information can be obtained. Due to its passive nature, passive sonar has good stealth capabilities and is suitable for long-range silent surveillance and long-term tracking tasks.
[0004] While both active and passive sonar have their advantages in underwater target tracking, they also have certain limitations in practical applications. Although active sonar can provide high-precision range and direction information, its reliance on actively emitting sound signals makes it less stealthy and easier to detect. Furthermore, it is susceptible to suppression or deception signals in environments with strong interference, leading to decreased tracking performance. In addition, its high-power transmission mechanism places high demands on the platform's energy consumption, and its detection accuracy is significantly reduced in long-range detection scenarios due to sound wave propagation attenuation.
[0005] In contrast, passive sonar possesses excellent concealment and strong anti-detection capabilities, making it suitable for long-range silent monitoring missions. However, it can only acquire target direction information and cannot directly measure distance parameters, thus presenting limitations in observational information. Furthermore, passive observation results are susceptible to environmental noise, multipath interference, and platform motion, leading to unstable observations and insufficient tracking performance. Especially in situations involving highly maneuverable targets or complex underwater acoustic environments, a single passive observation method struggles to achieve stable, continuous, and high-precision tracking.
[0006] In summary, a combined active and passive tracking method that integrates the advantages of both tracking methods has become a critical technical bottleneck that urgently needs to be overcome in this field. Summary of the Invention
[0007] This disclosure provides a method and system for joint active and passive tracking based on capacitive Kalman filtering.
[0008] Firstly, this disclosure provides a joint active-passive tracking method based on capacitive Kalman filtering, including:
[0009] A target state model is constructed for the tracking target based on the target state vector;
[0010] The observation models for active and passive sonars are constructed based on the target state model; wherein, the active and passive sonars receive acoustic wave data at their respective preset sampling periods.
[0011] The acoustic data received by the active sonar and the passive sonar are processed by capacitive Kalman filtering to obtain the estimated data of the active sonar and the passive sonar respectively. Among them, the estimated data is the target state vector estimation result and the covariance matrix of the target state estimation. The covariance matrix of the target state estimation characterizes the estimation accuracy of the target state vector estimation result.
[0012] By using an information fusion strategy based on variance fusion rules, the estimated data from active sonar and passive sonar are fused to obtain the fused target state vector estimation result, and then the target fused track is obtained.
[0013] Secondly, this disclosure provides a joint active-passive tracking system based on capacitive Kalman filtering, including:
[0014] The state model construction module is used to construct a target state model for the tracking target based on the target state vector.
[0015] The active sonar detection information acquisition-target tracking module is used to construct an active sonar observation model based on a target state model. The active sonar receives acoustic data at a preset sampling period and processes the received acoustic data through a capacitive Kalman filter to obtain active sonar estimation data. This active sonar estimation data includes the target state vector estimation result and the covariance matrix of the target state estimation. The covariance matrix of the active sonar target state estimation represents the estimation accuracy of the active sonar target state vector estimation result.
[0016] The passive sonar detection information acquisition-target tracking module is used to construct a passive sonar observation model based on the target state model. The passive sonar receives acoustic wave data at a preset sampling period. The acoustic wave data received by the passive sonar is processed by capacitive Kalman filtering to obtain passive sonar estimation data. Among them, the passive sonar estimation data consists of the target state vector estimation result of the passive sonar and the covariance matrix of the target state estimation. The covariance matrix of the target state estimation of the passive sonar represents the estimation accuracy of the target state vector estimation result of the passive sonar.
[0017] The data fusion module is used to fuse the estimated data from active sonar and passive sonar using an information fusion strategy designed based on variance fusion rules, to obtain the fused target state vector estimation result, and then to obtain the target fused track.
[0018] The content described in this section is not intended to identify key or important features of the embodiments of this disclosure, nor does it constitute a limitation on the scope of this disclosure.
[0019] Other features of this disclosure will be described in detail in the following description to aid understanding. Attached Figure Description
[0020] Figure 1 This disclosure provides an embodiment of a combined active and passive tracking method based on capacitive Kalman filtering;
[0021] Figure 2 This is a schematic diagram of the movement trajectory of the target and the observation station in one embodiment of this disclosure;
[0022] Figure 3 This is a schematic diagram of the individual tracking tracks of the active and passive sonars, the joint tracking track, and the actual movement track of the target in one embodiment of this disclosure;
[0023] Figure 4 This is a graph of RMSE over time in one embodiment of this disclosure;
[0024] Figure 5 This is an RMSE curve showing the variation of relative distance in one embodiment of this disclosure;
[0025] Figure 6This is a graph of RMSE over time in another embodiment of this disclosure;
[0026] Figure 7 This is an RMSE curve showing the variation with relative distance in another embodiment of this disclosure;
[0027] Figure 8 This is a schematic diagram of an active-passive joint tracking system based on capacitive Kalman filtering, provided in one embodiment of this disclosure. Detailed Implementation
[0028] The present disclosure will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the embodiments provided below are merely exemplary. Furthermore, for the sake of brevity and clarity, common knowledge has been omitted from the description of the embodiments below.
[0029] In this article, the terms "first," "second," and "third" are used only to distinguish the same or similar objects of description, and are not intended to limit the specific order or sequence of the objects being described, nor are they intended to limit the importance of the objects being described.
[0030] Figure 1 A flowchart of a joint active-passive tracking method based on capacitive Kalman filtering according to an embodiment of this disclosure is shown, which specifically includes:
[0031] Step S101: Construct a target state model for the tracking target based on the target state vector.
[0032] Step S102: Construct observation models for the active sonar and the passive sonar based on the target state model; wherein the active sonar and the passive sonar receive acoustic wave data at their respective preset sampling periods.
[0033] Step S103: Process the acoustic data received by the active sonar and the passive sonar respectively by capacitive Kalman filtering to obtain the estimation data of the active sonar and the passive sonar respectively; wherein, the estimation data is the target state vector estimation result obtained by the active sonar and the passive sonar at their respective sampling time and the covariance matrix of the target state estimation; the covariance matrix of the target state estimation characterizes the estimation accuracy of the target state vector estimation result.
[0034] Step S104: Using an information fusion strategy designed based on variance fusion rules, the estimated data from active sonar and passive sonar are fused to obtain the fused target state vector estimation result, and then the target fused track is obtained.
[0035] In one embodiment of this application, it is assumed that a target needs to be tracked in three dimensions using a combination of active and passive methods.
[0036] Construct the target state model:
[0037] Assuming the target moves at a constant velocity in a straight line, the target's state vector at time k is:
[0038]
[0039] in, The three-dimensional spatial position variable of the target; The three-dimensional spatial velocity variable is the target.
[0040] Its target state equation can be expressed as:
[0041]
[0042] in, The state transition matrix can be represented as:
[0043] The sampling interval;
[0044] With a mean of 0 and a covariance of Independent Gaussian process noise, used to characterize the disturbance caused by unknown maneuvers during the target's motion, can be expressed as:
[0045]
[0046] , , and and represent the power spectral density components of the process noise on the x, y, and z axes, respectively.
[0047] Constructing an active sonar observation model:
[0048] At time k, active sonar can measure the relative distance of the target with noise. Azimuth With pitch angle Information, the active sonar observation model can be represented as:
[0049]
[0050]
[0051] in, , and These represent the true values of the relative distance, azimuth, and elevation angles to the target as observed by active sonar, respectively. Location of the active sonar platform;
[0052] Observation noise All are Gaussian white noise with zero mean. This represents the nonlinear relationship between the target state and active sonar observations.
[0053] Constructing a passive sonar observation model:
[0054] At time k, the passive sonar can measure the azimuth of the target with noise. With pitch angle Information, the passive sonar observation model can be represented as:
[0055]
[0056]
[0057] in, and These represent the true values of the azimuth and elevation angles relative to the target as observed by passive sonar, respectively. Location of the passive sonar platform;
[0058] Observation noise All are Gaussian white noise with zero mean. This represents the nonlinear relationship between the target state and active sonar observations.
[0059] Both active and passive sonar systems process acoustic data using CKF.
[0060] For simplicity and ease of understanding, the observation models of the above active and passive sonar are uniformly abstracted into a nonlinear discrete system with a target state vector dimension of n:
[0061]
[0062] in: For state transition functions, for dimensional target state vector, For the observation vector, and These are uncorrelated process noise and observation noise.
[0063] CKF utilizes time updates and measurement updates to propagate the target state estimation vector and the covariance matrix of the target state estimation (used to characterize the estimation accuracy of the target state vector estimation result), thereby completing the real-time estimation of the target state variables. The specific steps are as follows:
[0064] Assuming the initial state, process noise, and observation noise are all uncorrelated, the target state estimation vector and the target state estimation covariance matrix need to be initialized first:
[0065]
[0066]
[0067] in, for The target state estimation vector at time t; For expectation operators; for The covariance matrix of the target state estimate at time t.
[0068] Time will then be updated:
[0069] right The covariance matrix of the target state estimate at time step 1 is subjected to Cholesky decomposition, and the result is as follows:
[0070]
[0071] Generate 2n volume points according to the following rules:
[0072]
[0073]
[0074]
[0075] in, express The i-th column.
[0076] Passing through the volume point Propagate to the predicted state to obtain the time-updated output:
[0077]
[0078]
[0079]
[0080] in, Let be the target prior state estimation vector at time k; Let be the covariance matrix of the prior state estimate of the target at time k.
[0081] The measurement was then updated:
[0082] Update the output based on time and regenerate the volume points:
[0083]
[0084] Pass the volume point Propagation yields updated measurement output:
[0085]
[0086]
[0087]
[0088]
[0089] in, Let k be the prior prediction observation vector of the target. Let be the covariance matrix of the prior predicted observations of the target at time k; Let be the cross-covariance matrix of the target prior state estimate and the target prior prediction observation at time k.
[0090] Finally, the state update result is obtained. After obtaining the new measurement update output, the target state estimation vector is updated according to the Kalman filter principle:
[0091]
[0092]
[0093]
[0094] in, Let K be the Kalman gain matrix at time k; Let be the target state estimation vector at time k; Let be the covariance matrix of the target state estimate at time k.
[0095] It should be understood that the above-mentioned CKF-based tracking method involves active and passive sonar operating independently. Passive sonar can only obtain the angle of arrival (in three-dimensional target observation, the angle of arrival refers to the azimuth and elevation angles), which is limited by the number of data dimensions, resulting in low target tracking accuracy. Although active sonar can obtain relative distance and angle of arrival, it needs to acquire acoustic data by transmitting signals and receiving their echoes. Therefore, its sampling period is usually longer than that of passive sonar, resulting in a lower detection data rate. At the same time, its acoustic data propagation path is two-way and there is loss, which also leads to low target tracking accuracy.
[0096] An information fusion strategy based on variance fusion rules is designed. Taking the passive sonar sampling time as the benchmark, the estimated data from both active and passive sonar are fused to obtain the fused target state vector estimation result, and thus the fused trajectory is obtained.
[0097] Since the sampling frequency of active sonar is lower than that of passive sonar, if only passive sonar samples at time k, the estimated result of the fused target state vector at that time is:
[0098]
[0099] If both active and passive sonar samples at time k, then the variance fusion rule is used to obtain the fused target state vector estimation result at that time:
[0100]
[0101] in, Let be the target state estimation vector obtained by the active sonar at time k; Let be the covariance matrix of the target state estimate obtained by the active sonar at time k; Let be the target state estimation vector obtained by the passive sonar at time k; Let be the covariance matrix of the target state estimate obtained by the passive sonar at time k.
[0102] This information fusion strategy allows the higher sampling rate of passive sonar detection information to compensate for the low data rate of active sonar detection, and the higher dimensionality of active sonar detection information to compensate for the low dimensionality of passive sonar detection information.
[0103] It should be noted that the active-passive joint tracking method based on capacitive Kalman filtering provided in this application adopts a distributed fusion strategy, which is reflected in the following two aspects:
[0104] Active and passive sonar operate independently and in parallel, respectively completing target tracking and observation. The active and passive sonars perform preliminary information extraction and state estimation locally, without the need to frequently upload raw data, thereby reducing communication bandwidth burden and improving response efficiency.
[0105] Considering the difference in sampling periods between active and passive sonar, and designing an information fusion strategy based on variance fusion rules, compared with a single sonar observation method, this approach can optimize the disadvantages of each in terms of observation dimension, sampling frequency, and acoustic data transmission loss, thereby improving the accuracy of target state estimation.
[0106] In one embodiment of this disclosure, the performance advantage of the joint active-passive tracking method over a single tracking method is verified through Monte Carlo simulation experiments. The simulation experiments use root mean squared error (RMSE) as the performance index, defined as:
[0107]
[0108]
[0109] in, and Let be the true position vector of the target at time k in the i-th Monte Carlo experiment and the estimated position vector of the algorithm, respectively. and In the i-th Monte Carlo experiment The actual velocity vector of the target at any given time and the estimated velocity vector of the algorithm, where M is the number of Monte Carlo simulation experiments, set to 2000.
[0110] Assume that the observation station (platform) is equipped with two types of sonar, and that the two sonars have the same preset sampling period.
[0111] Figure 2 The trajectory (track) of the target and the observation station (platform) according to an embodiment of this disclosure is shown. As the initial position, with The observation station moves at a constant velocity in a straight line from... Starting from the position, with The target moves at a constant linear speed, with a total continuous observation time of 100 seconds and a sampling interval of 1 second. Considering the performance limitations of active sonar at long distances, when the relative distance between the target and the detection platform is less than 800 meters, the arrival angle observation noise of both active and passive sonar is set to 1°, and the relative distance observation noise of active sonar is 3 meters. When the relative distance between the target and the detection platform is greater than 800 meters, the arrival angle observation noise of active sonar is adjusted to 3°, the relative distance observation noise is set to 5 meters, and the arrival angle observation noise of passive sonar remains at 1°.
[0112] Figure 3 , Figure 4 and Figure 5 The estimated trajectory (track) diagrams, RMSE curves at each tracking time (sampling time), and RMSE curves as a function of relative distance are shown in the scenarios described in this application embodiment when the active sonar operates independently, the passive sonar operates independently, and the active and passive sonar operate in combination.
[0113] In this embodiment, the relative distance between the target and the observation station gradually decreases from 1272m to 627.2m. Figure 4 , Figure 5 As shown, when the relative distance between the target and the observation station is greater than 800m, that is, within the first 60 seconds of tracking, the tracking results of the active sonar are significantly affected by observation noise and initial noise. Deviation from the true value and The inaccuracies of both active and passive sonar systems contribute to the initial noise, resulting in a significantly higher RMSE value compared to the subsequent 40 seconds. Passive sonar, limited by its observation dimensions, exhibits the highest tracking error throughout the entire tracking process. Compared to a single sonar system, the combined active and passive sonar scheme effectively integrates the advantages of both types of sonar: at long ranges, it mitigates the performance degradation caused by insufficient ranging accuracy of active sonar and the limited arrival angle information of passive sonar; at short ranges, its tracking accuracy is comparable to that of active sonar, while maintaining a low RMSE value.
[0114] In another embodiment of this application, simulations were performed for two operating conditions with different sonar sampling periods.
[0115] The sampling interval for the active sonar was set to 2 s, and the sampling interval for the passive sonar was set to 0.5 s. The remaining experimental parameters were consistent with those in the above embodiment. Figure 6 and Figure 7 This diagram shows the RMSE curves at various tracking moments and the RMSE curves as a function of relative distance for active sonar operating independently, passive sonar operating independently, and combined active and passive sonar operation in this scenario. (From...) Figure 6 and Figure 7 It is evident that higher-frequency passive sonar sampling significantly improves the combined active and passive tracking performance. In the first 60 seconds (when the relative distance between the target and the detection platform is greater than 800 m), Figure 6 and Figure 4 Compared to the previous results, denser sampling effectively mitigated the impact of limited active sonar performance on joint tracking performance in long-range detection, significantly accelerating the convergence speed and improving the accuracy of joint active and passive tracking; in the latter 40 s stage (relative distance less than 800 m), the joint scheme achieved higher tracking accuracy compared to the single active sonar scheme.
[0116] Figure 8 This application illustrates an embodiment of a joint active-passive tracking system based on capacitive Kalman filtering, specifically comprising:
[0117] The state model construction module 801 is used to construct a target state model for the tracking target based on the target state vector.
[0118] The active sonar detection information acquisition-target tracking module 802 is used to construct an active sonar observation model based on a target state model. The active sonar receives acoustic wave data at a preset sampling period and processes the received acoustic wave data through a capacitive Kalman filter to obtain active sonar estimation data. The active sonar estimation data includes the target state vector estimation result and the covariance matrix of the target state estimation. The covariance matrix of the target state estimation of the active sonar represents the estimation accuracy of the target state vector estimation result of the active sonar.
[0119] The passive sonar detection information acquisition-target tracking module 803 is used to construct a passive sonar observation model based on a target state model. The passive sonar receives acoustic wave data at a preset sampling period. The acoustic wave data received by the passive sonar is processed by a capacitive Kalman filter to obtain passive sonar estimation data. The passive sonar estimation data includes the target state vector estimation result of the passive sonar and the covariance matrix of the target state estimation. The covariance matrix of the target state estimation of the passive sonar represents the estimation accuracy of the target state vector estimation result of the passive sonar.
[0120] The data fusion module 804 is used to fuse active sonar estimation data and passive sonar estimation data based on variance fusion rules to obtain the fused target state vector estimation result, and then obtain the target fused track.
[0121] In another system embodiment of this application, the data fusion module 804 is further configured to:
[0122] If the active sonar samples synchronously at the sampling time of the passive sonar, then the variance fusion rule is used to fuse the estimated data of the active sonar and the passive sonar at that sampling time to obtain the fused target state vector estimation result at that sampling time.
[0123] If the active sonar does not sample at the sampling time of the passive sonar, then the fused target state vector estimation result at that sampling time is the target state vector estimation result of the passive sonar at that sampling time.
[0124] Any changes made to the above embodiments by those skilled in the art without departing from the true spirit and scope of this disclosure should be included within the scope of protection covered by the claims. The scope of protection claimed by this invention is limited only by the claims.
Claims
1. A joint active-passive tracking method based on capacitive Kalman filtering, characterized in that, include: A target state model is constructed for the tracking target based on the target state vector; Based on the target state model, observation models for active sonar and passive sonar are constructed respectively; wherein, the active sonar and passive sonar receive acoustic wave data at their respective preset sampling periods; The acoustic data received by the active sonar and the passive sonar are processed by capacitive Kalman filtering to obtain the estimation data of the active sonar and the passive sonar respectively; wherein, the estimation data is the target state vector estimation result and the covariance matrix of the target state estimation; the covariance matrix of the target state estimation characterizes the estimation accuracy of the target state vector estimation result; By using an information fusion strategy designed based on variance fusion rules, the estimated data of the active sonar and the passive sonar are fused to obtain the fused target state vector estimation result, and then the target fused track is obtained.
2. The method according to claim 1, characterized in that, Using an information fusion strategy designed based on variance fusion rules, the estimation data from the active sonar and passive sonar are fused to obtain a fused target state vector estimation result, including: If the active sonar samples synchronously at the sampling time of the passive sonar, then the estimated data of the active sonar and the passive sonar at the sampling time are fused using the variance fusion rule to obtain the fused target state vector estimation result at the sampling time. If the active sonar is not sampling at the sampling time of the passive sonar, then the fused target state vector estimation result at the sampling time is the target state vector estimation result of the passive sonar at the sampling time.
3. A combined active and passive tracking system based on capacitive Kalman filtering, characterized in that, include: The state model construction module is used to construct a target state model for the tracking target based on the target state vector. An active sonar detection information acquisition-target tracking module is used to construct an active sonar observation model based on the target state model. The active sonar receives acoustic wave data at a preset sampling period, and processes the received acoustic wave data through a capacitive Kalman filter to obtain active sonar estimation data. The active sonar estimation data includes the target state vector estimation result and the covariance matrix of the target state estimation. The covariance matrix of the target state estimation characterizes the estimation accuracy of the target state vector estimation result. A passive sonar detection information acquisition-target tracking module is used to construct a passive sonar observation model based on the target state model. The passive sonar receives acoustic wave data at a preset sampling period. The acoustic wave data received by the passive sonar is processed by a capacitive Kalman filter to obtain passive sonar estimation data. The passive sonar estimation data includes the target state vector estimation result of the passive sonar and the covariance matrix of the target state estimation. The covariance matrix of the target state estimation of the passive sonar represents the estimation accuracy of the target state vector estimation result of the passive sonar. The data fusion module is used to fuse the active sonar estimation data and the passive sonar estimation data using an information fusion strategy designed based on variance fusion rules, to obtain the fused target state vector estimation result, and then to obtain the target fused track.
4. The system according to claim 3, characterized in that, The data fusion module is also used for: If the active sonar samples synchronously at the sampling time of the passive sonar, then the estimated data of the active sonar and the passive sonar at the sampling time are fused using the variance fusion rule to obtain the fused target state vector estimation result at the sampling time. If the active sonar is not sampling at the sampling time of the passive sonar, then the fused target state vector estimation result at the sampling time is the target state vector estimation result of the passive sonar at the sampling time.