WO2020113353A1 - 一种机动目标的跟踪方法及系统 - Google Patents

一种机动目标的跟踪方法及系统 Download PDF

Info

Publication number
WO2020113353A1
WO2020113353A1 PCT/CN2018/118844 CN2018118844W WO2020113353A1 WO 2020113353 A1 WO2020113353 A1 WO 2020113353A1 CN 2018118844 W CN2018118844 W CN 2018118844W WO 2020113353 A1 WO2020113353 A1 WO 2020113353A1
Authority
WO
WIPO (PCT)
Prior art keywords
fuzzy
model
target
function
membership
Prior art date
Application number
PCT/CN2018/118844
Other languages
English (en)
French (fr)
Inventor
李良群
谢维信
刘宗香
Original Assignee
深圳大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳大学 filed Critical 深圳大学
Priority to PCT/CN2018/118844 priority Critical patent/WO2020113353A1/zh
Publication of WO2020113353A1 publication Critical patent/WO2020113353A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35—Clustering; Classification

Definitions

  • the invention relates to the technical field of target tracking, in particular to a tracking method and system of a maneuvering target.
  • the TS (full name Takagi–Sugeno) model is a fuzzy inference model proposed by Takagi and Sugeno. It can introduce fuzzy semantic information that can critically determine the motion model in a simple manner, and this model can approximate a nonlinear system of any shape .
  • EKF extended Kalman filter
  • a first aspect of the present invention provides a tracking method for a maneuvering target, including: estimating a state prediction value of a fuzzy linear model based on a TS fuzzy semantic model; estimating an estimated target state value of the fuzzy linear model based on the state prediction value; according to an observation model and The estimated target state value calculates the predicted observation value of each fuzzy model; calculates the fuzzy membership of the model of each fuzzy model based on the observation data set composed of the predicted observations; calculates the observed information of the target based on the fuzzy semantic model of the discrete dynamic system And heading angle error; integrate the observational information and heading angle error into the TS fuzzy model to update the antecedent parameters; calculate the antecedent parameter fuzzy membership of the antecedent parameters according to the model fuzzy membership; the fuzzy membership of the antecedent parameters
  • the composed fuzzy set and discrete dynamic system calculate the model weight of each linear model; calculate the target state value of the target according to the model weight, and calculate the target covariance of the target according to the target state value and the model weight; according to the target state value and The
  • a second aspect of the present invention provides a tracking system for a maneuvering target, including: a state prediction value module for estimating a state prediction value of a fuzzy linear model based on a TS fuzzy semantic model; a target state value module for predicting according to the state Value estimation The estimated target state value of the fuzzy linear model; The predicted observation value module, used to calculate the predicted observation value of each fuzzy model based on the observation model and the estimated target state value; The model fuzzy membership module, used to predict the observed value The formed observation data set is used to calculate the model fuzzy membership of each fuzzy model; the parameter module is used to calculate the observational innovation and heading angle error of the target according to the fuzzy semantic model of the discrete dynamic system; the updated predecessor parameter module is used to convert the observation The new information and heading angle error are integrated into the TS fuzzy model to update the antecedent parameters; the antecedent parameter fuzzy membership module is used to calculate the antecedent parameter fuzzy membership of the antecedent parameters based on the model fuzzy membership; the model weight module is used to Calculate the model weight of each linear model according to
  • a third aspect of the present invention provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer Program, implement any of the methods described above.
  • a fourth aspect of the present invention provides a computer-readable storage medium on which a computer program is stored.
  • the computer program is executed by a processor, any one of the methods described above is implemented.
  • the antecedent parameters can be updated, so that the antecedent parameters of the target can be more accurately identified, so that subsequent calculations can be obtained More accurate parameters of the antecedent, thereby improving the accuracy of the final calculation result, making the prediction of the trajectory of the maneuvering target more accurate.
  • FIG. 1 is a schematic flowchart of a method for tracking a maneuvering target according to an embodiment of the present invention
  • FIG. 2 is a schematic block diagram of a structure of an electronic device according to an embodiment of the present invention.
  • a first aspect of the present invention provides a tracking method for a maneuvering target, including: estimating a state prediction value of a fuzzy linear model based on a TS fuzzy semantic model; estimating an estimated target state value of the fuzzy linear model based on the state prediction value; according to an observation model and The estimated target state value calculates the predicted observation value of each fuzzy model; calculates the fuzzy membership of the model of each fuzzy model based on the observation data set composed of the predicted observations; calculates the observed information of the target based on the fuzzy semantic model of the discrete dynamic system And heading angle error; integrate the observational information and heading angle error into the TS fuzzy model to update the antecedent parameters; calculate the antecedent parameter fuzzy membership of the antecedent parameters according to the model fuzzy membership; the fuzzy membership of the antecedent parameters
  • the composed fuzzy set and discrete dynamic system calculate the model weight of each linear model; calculate the target state value of the target according to the model weight, and calculate the target covariance of the target according to the target state value and the model weight; according to the target state value and The
  • the estimation of the estimated target state value of the fuzzy linear model according to the state prediction value includes: introducing a least squares estimator; introducing the target speed and time interval of the target as the forgetting factor in the least squares estimator; The forgetting factor and least square estimator establish a modified extended forgetting factor least square estimator, and calculate the estimated target state value of the fuzzy linear model according to the modified extended forgetting factor least square estimator and the state prediction value.
  • the calculation of the model fuzzy membership of each fuzzy model based on the observation data set composed of predicted observations includes: setting cross entropy; setting fuzzy cross entropy according to the cross entropy; setting kernel fuzzy C based on fuzzy cross entropy Regression clustering function of the regression model clustering; calculate the fuzzy membership of each fuzzy model according to the regression clustering function and the observation data set.
  • the setting of the fuzzy cross entropy according to the cross entropy includes: setting the Gaussian function as the kernel function of the cross entropy; setting the sample mean estimation function of the cross entropy in the case of small samples; according to the sample mean estimation function and Fuzzy information processing theory, defining fuzzy cross entropy.
  • the setting of the regression clustering function of the kernel fuzzy C regression model clustering based on fuzzy cross entropy includes: setting an objective function of the kernel fuzzy C regression model clustering based on the observation data set and the output of the fuzzy model; setting The weighted exponent of the objective function and set the kernel space distance function; simplify the fuzzy cross entropy and define the modified objective function; bring the revised objective function into the kernel space distance function to obtain the fuzzy membership function, which is calculated according to the fuzzy membership function The fuzzy membership of each fuzzy model.
  • the updating of the antecedent parameters after incorporating the observational innovation and heading angle error into the TS fuzzy model includes: using three fixed-granularity fuzzy sets to describe the innovation and heading angle error, respectively; using a Gaussian membership function to represent the fixed granularity Fuzzy set; update the TS fuzzy model according to the antecedent parameters defined by the Gaussian membership function to obtain a modified TS fuzzy model; update the antecedent parameters according to the modified TS fuzzy model and the fuzzy membership of the model.
  • the method of setting the TS fuzzy model includes: setting a nonlinear function of a discrete nonlinear dynamic system; using a fuzzy linear model to represent the nonlinear function; obtaining a global fuzzy model according to the fuzzy linear model;
  • the fixed-bell membership function is a fuzzy membership function, and the fuzzy membership of the model in the global fuzzy model is calculated according to the bell membership function.
  • a second aspect of the present invention provides a tracking system for a maneuvering target, including: a state prediction value module for estimating a state prediction value of a fuzzy linear model based on a TS fuzzy semantic model; a target state value module for predicting according to the state Value estimation The estimated target state value of the fuzzy linear model; The predicted observation value module, used to calculate the predicted observation value of each fuzzy model based on the observation model and the estimated target state value; The model fuzzy membership module, used to predict the observed value The formed observation data set is used to calculate the model fuzzy membership of each fuzzy model; the parameter module is used to calculate the observational innovation and heading angle error of the target according to the fuzzy semantic model of the discrete dynamic system; the updated predecessor parameter module is used to convert the observation The new information and heading angle error are integrated into the TS fuzzy model to update the antecedent parameters; the antecedent parameter fuzzy membership module is used to calculate the antecedent parameter fuzzy membership of the antecedent parameters based on the model fuzzy membership; the model weight module is used to Calculate the model weight of each linear model according to
  • a third aspect of the present invention provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer Program, implement any of the methods described above.
  • a fourth aspect of the present invention provides a computer-readable storage medium on which a computer program is stored.
  • the computer program is executed by a processor, any one of the methods described above is implemented.
  • FIG. 1 for a tracking method of a maneuvering target, including: S1, estimating the state prediction value of the fuzzy linear model based on the TS fuzzy semantic model; S2, estimating the estimated target state value of the fuzzy linear model based on the state prediction value; S3, Calculate the predicted observations of each fuzzy model based on the observation model and estimated target state value; S4, calculate the model fuzzy membership of each fuzzy model based on the observation data set composed of the predicted observations; S5, based on the fuzzy semantic model of the discrete dynamic system Calculate the observed information and heading angle error of the target; S6. Integrate the observed information and heading angle error into the TS fuzzy model and update the antecedent parameters; S7.
  • the TS fuzzy model setting method includes: setting the nonlinear function of the discrete nonlinear dynamic system; using the fuzzy linear model to represent the nonlinear function; obtaining the global fuzzy model according to the fuzzy linear model; setting the bell-shaped membership function as the fuzzy membership Function, and calculate the fuzzy membership of the model in the global fuzzy model according to the bell membership function.
  • formula 1 and formula 2 represent discrete nonlinear dynamic systems, and formula 1 represents the following:
  • Equation 2 is expressed as follows:
  • Equation 1 and Equation 2 x k ⁇ R n represents the n-dimensional state vector at time k, z k ⁇ R m represents the m-dimensional observation vector, and f(x k-1 ) and h(x k ) represent suitable nonlinearities function.
  • e k-1 means the mean is 0 and the covariance is Process noise
  • v k means the mean is 0 and the covariance is Observation noise.
  • Equation 3 M fuzzy linear models as in Equation 3, which is expressed as follows:
  • Equation 4 the global fuzzy model
  • Equation 5 is expressed as follows:
  • Equation 6 Presentation variable Belongs to the model set Degree of membership
  • Equation 7 is expressed as follows:
  • Equation 7 With Respectively represent the mean and standard deviation of the jth membership function of the ith rule, so that the fuzzy membership of the antecedent parameters is calculated by Equation 7
  • Equation 3 the state of each model at k-1 can be set as Then the predicted state at time k-1 It can be expressed by formula 8, which is expressed as follows:
  • Estimating the estimated target state value of the fuzzy linear model based on the state prediction value includes: introducing the least square estimator; introducing the target speed and time interval of the target as the forgetting factor in the least square estimator; according to the forgetting factor and the least square estimator A modified extended forgetting factor least squares estimator is established, and the estimated target state value of the fuzzy linear model is calculated according to the modified extended forgetting factor least squares estimator and the state prediction value.
  • the speed v and the time interval of the target are introduced as the forgetting factor ⁇ ; under normal circumstances, the more accurate the current observation information, or When the historical data contains less information, the forgetting factor ⁇ is smaller, and conversely the forgetting factor is larger. Therefore, it can be known that the smaller the forgetting factor ⁇ is, the greater the forgetting factor is when the speed v is greater or the time interval is greater.
  • the least square estimator of the extended amnesia factor that is corrected is shown in Equation 9 to Equation 12:
  • Equation 9 is expressed as follows:
  • Equation 10 is expressed as follows:
  • Equation 11 is expressed as follows:
  • Equation 12 is expressed as follows:
  • Equation 9 to Equation 12 Represents the state estimate of model i at time k, Represents the state covariance of model i at time k, and w i,k-1 represents the weight of model i at time k-1, which is recorded as the model weight, and other variables are the same as Equation 3.
  • Calculating the model fuzzy membership of each fuzzy model based on the observation data set composed of predicted observations includes: setting cross entropy; setting fuzzy cross entropy according to cross entropy; setting kernel fuzzy C regression model clustering based on fuzzy cross entropy Regression clustering function; calculate the fuzzy membership of each fuzzy model according to the regression clustering function and the observation data set.
  • Setting fuzzy cross-entropy according to cross-entropy includes: setting the Gaussian function as the kernel function of cross-entropy; setting the sample mean estimation function of the cross entropy in the case of small samples; defining the fuzzy according to the sample mean estimation function and fuzzy information processing theory Cross entropy; identify the antecedent parameters of TS fuzzy model based on fuzzy cross entropy.
  • Cross entropy represents the generalized similarity measure between any two random variables, and is defined as Equation 13, which is expressed as follows:
  • Equation 13 the joint distribution function of F XY (x, y) random variables X and Y, E represents the mathematical expectation, and ⁇ ⁇ (X, Y) represents the shift-invariant Merer kernel.
  • the Gaussian kernel function is selected as the kernel function of cross entropy, then ⁇ ⁇ (X, Y) is expressed as Equation 14, and Equation 14 is expressed as follows:
  • Equation 14 ⁇ represents the core size; the joint distribution function of X and Y is unknown.
  • Equation 15 the sample mean estimate of cross entropy is defined as Equation 15, which is expressed as follows:
  • Equation 16 m is the weighted index, and ⁇ i represents the fuzzy membership between the variables x i and y i , and satisfies Equation 17, which is expressed as follows:
  • Setting the regression clustering function of the kernel fuzzy C regression model clustering based on fuzzy cross entropy includes: setting the objective function of the kernel fuzzy C regression model clustering based on the observation data set and the output of the fuzzy model; setting the weighted index of the objective function , And set the kernel space distance function; simplify the fuzzy cross entropy, and define the modified objective function, and identify the posterior parameters of the TS fuzzy model according to the revised objective function; bring the revised objective function into the kernel space distance function, and according to the antecedent parameters
  • the fuzzy membership function is obtained based on the following parameters, and the fuzzy membership of each fuzzy model is calculated according to the fuzzy membership function.
  • Equation 18 Equation 18
  • Equation 19 is expressed as follows:
  • D ij represents observation
  • fuzzy model output The measure of dissimilarity between, here, D ij is defined as the nuclear space distance, and the specific expression of D ij is as shown in formula 20, which is expressed as follows:
  • ⁇ represents any non-linear mapping from the original feature space to the high-dimensional feature space
  • K( ⁇ ) represents the Mercer kernel function
  • equation 16 is reduced to equation 21
  • formula 21 means as follows:
  • formula 22 is defined as the modified objective function L k , and formula 22 is expressed as follows:
  • Equation 22 ⁇ is the Lagrangian multiplier vector, combining Equation 11 and Equation 14, we get Equation 23, which is expressed as follows:
  • observational innovation and heading angle error into the TS fuzzy model and updating of the antecedent parameters include: using three fixed-grain fuzzy sets to describe the new information and heading angle errors; using a Gaussian membership function to represent the fixed-grain fuzzy set; according to Gaussian membership
  • the predecessor parameters defined by the function update the TS fuzzy model to obtain the modified TS fuzzy model; the predecessor parameters are updated according to the modified TS fuzzy model and model fuzzy membership.
  • formula 3 in maneuvering target tracking, select the observation new information ⁇ v k and the heading angle error As the antecedent variable of TS fuzzy model. Assuming observation z k at time k , Represents the target state at time k-1, then ⁇ v k and Can be defined separately as formula 28 and formula 29, formula 28 is expressed as follows:
  • Equation 29 is expressed as follows:
  • Equation 30 is expressed as follows:
  • Equation 29 to Equation 30 ⁇ v k represents the observed innovation, Indicates the heading angle error, Represents the target heading angle at time k, Represents the predicted observation at time k, with Represent the target state vector The x and y components in.
  • Equation 32 is expressed as follows:
  • Equation 31 and Equation 32 with Represent the mean and variance of the jth linguistic value of innovation at time k, with Respectively represent the mean and variance of the jth linguistic value of heading angle error at time k.
  • M represents the rule number, with Respectively represent the state transition matrix and the observation matrix, ⁇ i represents the target turning rate.
  • Equation 29 and Equation 30 need to be updated at all times, and the fuzzy membership u i,k obtained by clustering according to Equation 23, the parameters of the antecedent variables can be updated as Equation 33 to Equation 36.
  • 33 means as follows:
  • Equation 34 is expressed as follows:
  • Equation 35 is expressed as follows:
  • Equation 36 is expressed as follows:
  • Equation 37 the target state value can be obtained and target covariance P k , target state value
  • Equation 38 the target covariance P k is shown in Equation 38, and Equation 37 is expressed as follows:
  • Equation 38 is expressed as follows:
  • An embodiment of the present application provides a tracking system for a maneuvering target, including: a state prediction value module for estimating a state prediction value of a fuzzy linear model based on a TS fuzzy semantic model; a target state value module for estimating fuzzy linearity based on the state prediction value The estimated target state value of the model; the predicted observation value module, which is used to calculate the predicted observation value of each fuzzy model based on the observation model and the estimated target state value; the model fuzzy membership module, which is used to calculate the observation data set composed of the predicted observation value The fuzzy membership of each fuzzy model; the parameter module, which is used to calculate the observed innovation and heading angle error of the target based on the fuzzy semantic model of the discrete dynamic system; the updated predecessor parameter module, which is used to convert the observed innovation and heading angle error
  • the antecedent parameters are updated after being integrated into the TS fuzzy model; the antecedent parameter fuzzy membership module is used to calculate the antecedent parameter fuzzy membership of the antecedent parameters according to the model fuzzy membership; the model weight module is used to blur the antecedent
  • An embodiment of the present application provides an electronic device. Please refer to 2.
  • the electronic device includes a memory 601, a processor 602, and a computer program stored on the memory 601 and executable on the processor 602.
  • the processor 602 executes the computer program At this time, the tracking method of the maneuvering target described in the foregoing embodiment is realized.
  • the electronic device further includes: at least one input device 603 and at least one output device 604.
  • the aforementioned memory 601, processor 602, input device 603, and output device 604 are connected via a bus 605.
  • the input device 603 may specifically be a camera, a touch panel, a physical button, a mouse, or the like.
  • the output device 604 may specifically be a display screen.
  • the memory 601 may be a high-speed random access memory (RAM, Random Access Memory) memory, or may be a non-volatile memory (non-volatile memory), such as a disk memory.
  • RAM Random Access Memory
  • non-volatile memory non-volatile memory
  • the memory 601 is used to store a set of executable program codes, and the processor 602 is coupled to the memory 601.
  • the embodiments of the present application further provide a computer-readable storage medium.
  • the computer-readable storage medium may be provided in the electronic device in the foregoing embodiments, and the computer-readable storage medium may be the foregoing embodiment ⁇ 601.
  • a computer program is stored on the computer-readable storage medium, and when the program is executed by the processor 602, the tracking method of the maneuvering target described in the foregoing method embodiments is implemented.
  • the computer storable medium may also be various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory 601 (ROM, Read-Only Memory), RAM, a magnetic disk, or an optical disk.
  • program codes such as a U disk, a mobile hard disk, a read-only memory 601 (ROM, Read-Only Memory), RAM, a magnetic disk, or an optical disk.
  • the disclosed device and method may be implemented in other ways.
  • the device embodiments described above are only schematic.
  • the division of the modules is only a division of logical functions.
  • there may be other divisions for example, multiple modules or components may be combined or Can be integrated into another system, or some features can be ignored, or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or modules, and may be in electrical, mechanical, or other forms.
  • modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist alone physically, or two or more modules may be integrated into one module.
  • the above integrated modules may be implemented in the form of hardware or software function modules.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

一种机动目标的跟踪方法及系统，用于机动目标跟踪，解决了现有技术中对机动目标的准确估计上，估计结果与实际情况之间仍存在较大误差的问题，其包括：基于T-S模糊语义模型估计模糊线性模型的状态预测值（S1）；估计模糊线性模型的估计目标状态值计算每个模糊模型的预测观测值（S3）；计算每个模糊模型的模型模糊隶属度（S4）；计算目标的观测新息及航向角误差（S5）；将观测新息及航向角误差融入T-S模糊模型后更新前件参数（S6）；计算前件参数的前件参数模糊隶属度（S7）；计算每个线性模型的模型权值（S8）；计算目标的目标状态值，并计算目标的目标协方差（S9）；估计机动目标的运动轨迹（S10）；从而降低了估计结果与实际情况之间的误差。

Description

一种机动目标的跟踪方法及系统 技术领域
本发明涉及目标跟踪技术领域，尤其涉及一种机动目标的跟踪方法及系统。
背景技术
T-S(全称为Takagi–Sugeno)模型是Takagi和Sugeno提出的一种模糊推理模型，它能够以简单的方式引入能够关键性决定运动模型的模糊语义信息，并且这个模型可以逼近任意形状的非线性系统。
当前研究的机动目标跟踪算法大多是基于概率统计模型，这很难满足系统性能的需求。扩展卡尔曼滤波器(EKF)使用泰勒公式线性化测量和状态模型去近似当前的状态估计，卡尔曼滤波(KF)算法的迭代仍然得到应用。可是当目标做强机动时，EKF的性能下降地非常迅速，且算法还会发散。
针对上述问题，Julier和Uhlmann提出了无迹卡尔曼滤波(UKF)算法，UKF算法在二价泰勒展开时，通过迭代地传播一些精确的sigma点来获取后验均值和协方差。
技术问题
根据UKF算法进行研究的成果依然很难满足实际应用对非线性非高斯系统状态估计所提出的鲁棒性和准确性的要求，因此在对机动目标的准确估计上，估计结果与实际情况之间仍存在较大误差。
技术解决方案
本发明第一方面提供一种机动目标的跟踪方法，包括：基于T-S模糊语义模型估计模糊线性模型的状态预测值；根据所述状态预测值估计模糊线性模型的估计目标状态值；根据观测模型及所述估计目标状态值计算每个模糊模型的预测观测值；根据预测观测值组成的观测数据集计算每个模糊模型的模型模糊隶属度；根据离散动态系统的模 糊语义模型计算目标的观测新息及航向角误差；将观测新息及航向角误差融入T-S模糊模型后更新前件参数；根据模型模糊隶属度计算前件参数的前件参数模糊隶属度；根据所述前件参数的模糊隶属度组成的模糊集及离散动态系统计算每个线性模型的模型权值；根据模型权值计算目标的目标状态值，并根据目标状态值及模型权值计算目标的目标协方差；根据目标状态值及目标协方差估计机动目标的运动轨迹。
本发明第二方面提供了一种机动目标的跟踪系统，包括：状态预测值模块，用于基于T-S模糊语义模型估计模糊线性模型的状态预测值；目标状态值模块，用于根据所述状态预测值估计模糊线性模型的估计目标状态值；预测观测值模块，用于根据观测模型及所述估计目标状态值计算每个模糊模型的预测观测值；模型模糊隶属度模块，用于根据预测观测值组成的观测数据集计算每个模糊模型的模型模糊隶属度；参数模块，用于根据离散动态系统的模糊语义模型计算目标的观测新息及航向角误差；更新前件参数模块，用于将观测新息及航向角误差融入T-S模糊模型后更新前件参数；前件参数模糊隶属度模块，用于根据模型模糊隶属度计算前件参数的前件参数模糊隶属度；模型权值模块，用于根据所述前件参数模糊隶属度组成的模糊集及离散动态系统计算每个线性模型的模型权值；目标协方差模块，用于根据模型权值计算目标的目标状态值，并根据目标状态值及模型权值计算目标的目标协方差；运动轨迹估计模块，用于根据目标状态值及目标协方差估计机动目标的运动轨迹。
本发明第三方面提供了一种电子装置，包括：存储器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机程序，其特征在于，所述处理器执行所述计算机程序时，实现上述中的任意一项所 述方法。
本发明第四方面提供了一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时，实现上述中的任意一项所述方法。
有益效果
通过计算每个模糊模型的预测观测值，并将观测新息及航向角误差融入T-S模糊模型，能够更新前件参数，从而对目标的前件参数能够更加准确地辨识，使得后续的计算能够得到更加精准的前件参数，从而提高了最终计算结果的准确性，使得对机动目标的运动轨迹的预测更加精准。
附图说明
图1为本发明实施例机动目标的跟踪方法的流程示意框图；
图2为本发明实施例电子装置的结构示意框图。
本发明的最佳实施方式
本发明第一方面提供一种机动目标的跟踪方法，包括：基于T-S模糊语义模型估计模糊线性模型的状态预测值；根据所述状态预测值估计模糊线性模型的估计目标状态值；根据观测模型及所述估计目标状态值计算每个模糊模型的预测观测值；根据预测观测值组成的观测数据集计算每个模糊模型的模型模糊隶属度；根据离散动态系统的模糊语义模型计算目标的观测新息及航向角误差；将观测新息及航向角误差融入T-S模糊模型后更新前件参数；根据模型模糊隶属度计算前件参数的前件参数模糊隶属度；根据所述前件参数的模糊隶属度组成的模糊集及离散动态系统计算每个线性模型的模型权值；根据模型权值计算目标的目标状态值，并根据目标状态值及模型权值计算目标的目标协方差；根据目标状态值及目标协方差估计机动目标的运动轨迹。
进一步地，所述根据所述状态预测值估计模糊线性模型的估计目 标状态值包括：引入最小二乘估计器；引入目标的目标速度和时间间隔作为最小二乘估计器内的遗忘因子；根据所述遗忘因子及最小二乘估计器建立修正的扩展遗忘因子最小二乘估计器，并根据修正的扩展遗忘因子最小二乘估计器及所述状态预测值计算模糊线性模型的估计目标状态值。
进一步地，所述根据预测观测值组成的观测数据集计算每个模糊模型的模型模糊隶属度包括：设定交叉熵；根据交叉熵设定模糊交叉熵；设定基于模糊交叉熵的核模糊C回归模型聚类的回归聚类函数；根据所述回归聚类函数及所述观测数据集计算每个模糊模型的模糊隶属度。
进一步地，所述根据交叉熵设定模糊交叉熵包括：设定高斯函数为交叉熵的核函数；设定在小样本情况下，交叉熵的样本均值估计函数；根据所述样本均值估计函数及模糊信息处理理论，定义模糊交叉熵。
进一步地，所述设定基于模糊交叉熵的核模糊C回归模型聚类的回归聚类函数包括：根据观测数据集及模糊模型的输出设定核模糊C回归模型聚类的目标函数；设定目标函数的加权指数，并设定核空间距离函数；简化所述模糊交叉熵，并定义修正目标函数；将修正目标函数带入核空间距离函数，得到模糊隶属度函数，根据模糊隶属度函数计算每个模糊模型的模型模糊隶属度。
进一步地，所述将观测新息及航向角误差融入T-S模糊模型后更新前件参数包括：分别采用三个固定粒度的模糊集描述新息及航向角误差；使用高斯隶属函数表示所述固定粒度的模糊集；根据所述高斯隶属函数定义的前件参数更新T-S模糊模型，得到修改T-S模糊模型； 根据修改T-S模糊模型及所述模型模糊隶属度更新前件参数。
进一步地，所述T-S模糊模型的设定方法包括：设定离散非线性动态系统的非线性函数；使用模糊线性模型表示所述非线性函数；根据所述模糊线性模型得出全局模糊模型；设定钟型隶属函数为模糊隶属度函数，并根据所述钟型隶属度函数计算所述全局模糊模型内模型的模糊隶属度。
本发明第二方面提供了一种机动目标的跟踪系统，包括：状态预测值模块，用于基于T-S模糊语义模型估计模糊线性模型的状态预测值；目标状态值模块，用于根据所述状态预测值估计模糊线性模型的估计目标状态值；预测观测值模块，用于根据观测模型及所述估计目标状态值计算每个模糊模型的预测观测值；模型模糊隶属度模块，用于根据预测观测值组成的观测数据集计算每个模糊模型的模型模糊隶属度；参数模块，用于根据离散动态系统的模糊语义模型计算目标的观测新息及航向角误差；更新前件参数模块，用于将观测新息及航向角误差融入T-S模糊模型后更新前件参数；前件参数模糊隶属度模块，用于根据模型模糊隶属度计算前件参数的前件参数模糊隶属度；模型权值模块，用于根据所述前件参数模糊隶属度组成的模糊集及离散动态系统计算每个线性模型的模型权值；目标协方差模块，用于根据模型权值计算目标的目标状态值，并根据目标状态值及模型权值计算目标的目标协方差；运动轨迹估计模块，用于根据目标状态值及目标协方差估计机动目标的运动轨迹。
本发明第三方面提供了一种电子装置，包括：存储器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机程序，其特征在于，所述处理器执行所述计算机程序时，实现上述中的任意一项所 述方法。
本发明第四方面提供了一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时，实现上述中的任意一项所述方法。
本发明的实施方式
为使得本发明的发明目的、特征、优点能够更加的明显和易懂，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而非全部实施例。基于本发明中的实施例，本领域技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
请参阅图1，为一种机动目标的跟踪方法，包括：S1、基于T-S模糊语义模型估计模糊线性模型的状态预测值；S2、根据状态预测值估计模糊线性模型的估计目标状态值；S3、根据观测模型及估计目标状态值计算每个模糊模型的预测观测值；S4、根据预测观测值组成的观测数据集计算每个模糊模型的模型模糊隶属度；S5、根据离散动态系统的模糊语义模型计算目标的观测新息及航向角误差；S6、将观测新息及航向角误差融入T-S模糊模型后更新前件参数；S7、根据模型模糊隶属度计算前件参数的前件参数模糊隶属度；S8、根据前件参数模糊隶属度组成的模糊集及离散动态系统计算每个线性模型的模型权值；S9、根据模型权值计算目标的目标状态值，并根据目标状态值及模型权值计算目标的目标协方差；S10、根据目标状态值及目标协方差估计机动目标的运动轨迹。
T-S模糊模型的设定方法包括：设定离散非线性动态系统的非线性函数；使用模糊线性模型表示非线性函数；根据模糊线性模型得出 全局模糊模型；设定钟型隶属函数为模糊隶属度函数，并根据钟型隶属度函数计算全局模糊模型内模型的模糊隶属度。
具体地，设定公式1及公式2表示离散非线性动态系统，公式1表示如下：
x k＝f(x k-1)+e k-1
公式2表示如下：
z k＝h(x k)+v k
在公式1及公式2中，x k∈R n表示k时刻n维状态矢量，z k∈R m表示m维观测矢量，f(x k-1)和h(x k)表示合适的非线性函数。e k-1表示均值为0协方差为
Figure PCTCN2018118844-appb-000001
的过程噪声，v k表示均值为0协方差为
Figure PCTCN2018118844-appb-000002
的观测噪声。
在本领域中，T-S模糊模型认为任何非线性系统可以用如公式3中的M个模糊线性模型表示，公式3表示如下：
Figure PCTCN2018118844-appb-000003
其中，θ k表示规则的前件变量，
Figure PCTCN2018118844-appb-000004
表示前件变量对应的模糊隶属函，
Figure PCTCN2018118844-appb-000005
和
Figure PCTCN2018118844-appb-000006
分别表示状态转移矩阵和观测矩阵；从公式3可以看出，M模糊模型都是线性时不变模型；于是，全局模糊模型可以如公式4及公式5所示，公式4表示如下：
Figure PCTCN2018118844-appb-000007
公式5表示如下：
Figure PCTCN2018118844-appb-000008
在公式4及公式5中，
Figure PCTCN2018118844-appb-000009
表示x k属于第i个线性模型的模糊隶属度，可以通过公式6计算，公式6表示如下：
Figure PCTCN2018118844-appb-000010
在公式6中，
Figure PCTCN2018118844-appb-000011
表示变量
Figure PCTCN2018118844-appb-000012
属于模型集
Figure PCTCN2018118844-appb-000013
的隶属度，且
Figure PCTCN2018118844-appb-000014
模糊隶属函数
Figure PCTCN2018118844-appb-000015
采用如公式7所示的钟型隶属函数，公式7表示如下：
Figure PCTCN2018118844-appb-000016
在公式7中，
Figure PCTCN2018118844-appb-000017
和
Figure PCTCN2018118844-appb-000018
分别表示第i个规则第j个隶属度函数的均值和标准差，从而通过公式7计算前件参数模糊隶属度
Figure PCTCN2018118844-appb-000019
根据公式3可设定每个模型在k-1时刻的状态为
Figure PCTCN2018118844-appb-000020
则k-1时刻的预测状态
Figure PCTCN2018118844-appb-000021
可用公式8表示，公式8表示如下：
Figure PCTCN2018118844-appb-000022
根据状态预测值估计模糊线性模型的估计目标状态值包括：引入最小二乘估计器；引入目标的目标速度和时间间隔作为最小二乘估计器内的遗忘因子；根据遗忘因子及最小二乘估计器建立修正的扩展遗忘因子最小二乘估计器，并根据修正的扩展遗忘因子最小二乘估计器及状态预测值计算模糊线性模型的估计目标状态值。
为了提高表示T-S模糊模型的后件参数的准确性，在引入最小二乘估计器的基础上，引入目标的速度v与时间间隔作为遗忘因子λ；在通常情况下，当前观测信息越精确，或历史数据包含的信息越少时，遗忘因子λ越小，反之遗忘因子越大，因此可以得知，在速度v越大或时间间隔越大的情况下，遗忘因子λ越小，反之遗忘因子越大，因此得到修正的扩展遗忘因子最小二乘估计器如公式9至公式12所示：
公式9表示如下：
Figure PCTCN2018118844-appb-000023
公式10表示如下：
Figure PCTCN2018118844-appb-000024
公式11表示如下：
Figure PCTCN2018118844-appb-000025
公式12表示如下：
Figure PCTCN2018118844-appb-000026
在公式9至公式12中，
Figure PCTCN2018118844-appb-000027
表示k时刻模型i的状态估计值，
Figure PCTCN2018118844-appb-000028
表示k时刻模型i的状态协方差，w i,k-1表示表示k-1时刻模型i的权值，记做模型权值，其他变量与公式3相同。
根据预测观测值组成的观测数据集计算每个模糊模型的模型模糊隶属度包括：设定交叉熵；根据交叉熵设定模糊交叉熵；设定基于模糊交叉熵的核模糊C回归模型聚类的回归聚类函数；根据回归聚类函数及观测数据集计算每个模糊模型的模糊隶属度。
根据交叉熵设定模糊交叉熵包括：设定高斯函数为交叉熵的核函数；设定在小样本情况下，交叉熵的样本均值估计函数；根据样本均值估计函数及模糊信息处理理论，定义模糊交叉熵；根据模糊交叉熵辨识T-S模糊模型的前件参数。
交叉熵表示任意两个随机变量之间的广义相似测度，定义为公式13，公式13表示如下：
V σ(X,Y)＝E[κ σ(X,Y)]＝∫κ σ(X,Y)dF XY(x,y)
在公式13中，F XY(x,y)随机变量X和Y的联合分布函数，E表示数学期望，κ σ(X,Y)表示移不变Merer核。在本实施例中，选择高斯 核函数作为交叉熵的核函数，则κ σ(X,Y)表示如公式14，公式14表示如下：
Figure PCTCN2018118844-appb-000029
在公式14中，σ表示核尺寸；X和Y的联合分布函数未知。在小样本情况，交叉熵的样本均值估计定义如公式15，公式15表示如下：
Figure PCTCN2018118844-appb-000030
在公式15中，N表示数据对(x i,y i)的个数。从上式交叉熵的定义可以看出，对于所有样本都具有相同的权值1/N。而在实际中，不同样本对于状态估计的作用应该是不竟相同的，及不同样本应该具有不同的权值。对此，基于模糊信息处理理论，定义模糊交叉熵如公式16，公式16表示如下：
Figure PCTCN2018118844-appb-000031
在公式16中，m为加权指数，μ i表示变量x i和y i之间的模糊隶属度，且满足公式17，公式17表示如下：
Figure PCTCN2018118844-appb-000032
从公式11可知，当m等于0时，模糊交叉熵就退化成普通交叉熵。
设定基于模糊交叉熵的核模糊C回归模型聚类的回归聚类函数包括：根据观测数据集及模糊模型的输出设定核模糊C回归模型聚类的目标函数；设定目标函数的加权指数，并设定核空间距离函数；简化模糊交叉熵，并定义修正目标函数，且根据修正目标函数辨识T-S模糊模型的后件参数；将修正目标函数带入核空间距离函数，并根据 前件参数及后件参数得到模糊隶属度函数，根据模糊隶属度函数计算每个模糊模型的模型模糊隶属度。
设定在k时刻，总共接收到N个观测数据集
Figure PCTCN2018118844-appb-000033
同时又M个模糊输出
Figure PCTCN2018118844-appb-000034
聚类的目标就是将数据集Z k分成M类，并优化出观测与线性模型输出之间的隶属度矩阵U＝[u ij] M×N。u ij表示观测
Figure PCTCN2018118844-appb-000035
输入第i类的模糊隶属度。则核模糊C回归模型聚类的目标函数可以定义如公式18及公式19，公式18表示如下：
Figure PCTCN2018118844-appb-000036
公式19表示如下：
Figure PCTCN2018118844-appb-000037
在公式18及公式19中，m∈[1,∞]为加权指数，设为m＝2，D ij表示观测
Figure PCTCN2018118844-appb-000038
和模糊模型输出
Figure PCTCN2018118844-appb-000039
之间的相异性测度，这里，D ij定义为核空间距离，且D ij的具体表示如公式20，公式20表示如下：
Figure PCTCN2018118844-appb-000040
在公式20中，φ表示原始特征空间到高维特征空间的任意非线性映射，K(·)表示Mercer核函数，如果K(·)选用高斯核函数，则公式16化简为公式21，公式21表示如下：
Figure PCTCN2018118844-appb-000041
而为了引入模糊交叉熵，定义公式22为修正目标函数L k，公式22表示如下：
L k＝V F,σ-β·J
在公式22中，β为拉格朗日乘子矢量，联合公式11及公式14，得到公式23，公式23表示如下：
Figure PCTCN2018118844-appb-000042
为了计算出u ij，对u ij求一阶导数并令其等于0，可得公式24，公式24表示如下：
Figure PCTCN2018118844-appb-000043
更进一步可得公式25，公式25表示如下：
Figure PCTCN2018118844-appb-000044
使用公式21代替公式15中的u ij，并化简得到公式26，公式26表示如下：
Figure PCTCN2018118844-appb-000045
将公式22代入公式21，得到模糊隶属度u ij为公式27，公式27表示如下：
Figure PCTCN2018118844-appb-000046
将观测新息及航向角误差融入T-S模糊模型后更新前件参数包括：分别采用三个固定粒度的模糊集描述新息及航向角误差；使用高斯隶属函数表示固定粒度的模糊集；根据高斯隶属函数定义的前件参数更新T-S模糊模型，得到修改T-S模糊模型；根据修改T-S模糊模型及模型模糊隶属度更新前件参数。
根据公式3，在机动目标跟踪中，选择观测新息Δv k和航向角误 差
Figure PCTCN2018118844-appb-000047
作为T-S模糊模型的前件变量。假设k时刻的观测z k，
Figure PCTCN2018118844-appb-000048
表示k-1时刻的目标状态，则Δv k和
Figure PCTCN2018118844-appb-000049
可以分别定义如公式28及公式29，公式28表示如下：
Figure PCTCN2018118844-appb-000050
公式29表示如下：
Figure PCTCN2018118844-appb-000051
其中，
Figure PCTCN2018118844-appb-000052
如公式30所示，公式30表示如下：
Figure PCTCN2018118844-appb-000053
在公式29至公式30中，Δv k表示观测的新息，
Figure PCTCN2018118844-appb-000054
表示航向角误差，
Figure PCTCN2018118844-appb-000055
表示k时刻的目标航向角，
Figure PCTCN2018118844-appb-000056
表示k时刻的预测观测，
Figure PCTCN2018118844-appb-000057
和
Figure PCTCN2018118844-appb-000058
分别表示目标状态向量
Figure PCTCN2018118844-appb-000059
中的x分量和y分量。
而为了将上述信息融入T-S模糊模型中，采用几个固定粒度的模糊集分别描述新息和航向角误差；在T-S模糊模型中，新息变量Δv k利用三个语言值Small(S),Medium(M),and Large(L))描述，分别表示为
Figure PCTCN2018118844-appb-000060
和
Figure PCTCN2018118844-appb-000061
航向角误差
Figure PCTCN2018118844-appb-000062
使用三个语言值Negative Large(NL),Small(S)和Positive Large(PL)描述，分别表示为
Figure PCTCN2018118844-appb-000063
Figure PCTCN2018118844-appb-000064
和
Figure PCTCN2018118844-appb-000065
同时也假设都采用高斯隶属函数来表示上述模糊集，则可以设定公式31及公式32，公式31表示如下：
Figure PCTCN2018118844-appb-000066
公式32表示如下：
Figure PCTCN2018118844-appb-000067
在公式31及公式32中，
Figure PCTCN2018118844-appb-000068
和
Figure PCTCN2018118844-appb-000069
分别表示k时刻新息第j个语 言值的均值和方差，
Figure PCTCN2018118844-appb-000070
和
Figure PCTCN2018118844-appb-000071
分别表示k时刻航向角误差第j个语言值的均值和方差。
则根据上述设定的前件变量，公式3中的T-S模糊模型修改如下：
Figure PCTCN2018118844-appb-000072
其中，M表示规则数，
Figure PCTCN2018118844-appb-000073
和
Figure PCTCN2018118844-appb-000074
分别表示状态转移矩阵和观测矩阵，ω i表示目标转弯率。
在机动目标中，需要时刻对公式29和公式30中的参数进行更新，根据公式23聚类得到的模糊隶属度u i,k，前件变量的参数可以如公式33至公式36进行更新，公式33表示如下：
Figure PCTCN2018118844-appb-000075
公式34表示如下：
Figure PCTCN2018118844-appb-000076
公式35表示如下：
Figure PCTCN2018118844-appb-000077
公式36表示如下：
Figure PCTCN2018118844-appb-000078
根据上述的计算，可以得到目标状态值
Figure PCTCN2018118844-appb-000079
及目标协方差P k，目标状态值
Figure PCTCN2018118844-appb-000080
如公式37所示，目标协方差P k如公式38所示，公式37表示如下：
Figure PCTCN2018118844-appb-000081
公式38表示如下：
Figure PCTCN2018118844-appb-000082
最后根据目标状态值
Figure PCTCN2018118844-appb-000083
及目标协方差P k估计机动目标的运动轨 迹。
本申请实施例提供一种机动目标的跟踪系统，包括：状态预测值模块，用于基于T-S模糊语义模型估计模糊线性模型的状态预测值；目标状态值模块，用于根据状态预测值估计模糊线性模型的估计目标状态值；预测观测值模块，用于根据观测模型及估计目标状态值计算每个模糊模型的预测观测值；模型模糊隶属度模块，用于根据预测观测值组成的观测数据集计算每个模糊模型的模型模糊隶属度；参数模块，用于根据离散动态系统的模糊语义模型计算目标的观测新息及航向角误差；更新前件参数模块，用于将观测新息及航向角误差融入T-S模糊模型后更新前件参数；前件参数模糊隶属度模块，用于根据模型模糊隶属度计算前件参数的前件参数模糊隶属度；模型权值模块，用于根据前件参数的模糊隶属度组成的模糊集及离散动态系统计算每个线性模型的模型权值；目标协方差模块，用于根据模型权值计算目标的目标状态值，并根据目标状态值及模型权值计算目标的目标协方差；运动轨迹估计模块，用于根据目标状态值及目标协方差估计机动目标的运动轨迹。
本申请实施例提供一种电子装置，请参阅2，该电子装置包括：存储器601、处理器602及存储在存储器601上并可在处理器602上运行的计算机程序，处理器602执行该计算机程序时，实现前述实施例中描述的机动目标的跟踪方法。
进一步的，该电子装置还包括：至少一个输入设备603以及至少一个输出设备604。
上述存储器601、处理器602、输入设备603以及输出设备604，通过总线605连接。
其中，输入设备603具体可为摄像头、触控面板、物理按键或者鼠标等等。输出设备604具体可为显示屏。
存储器601可以是高速随机存取记忆体(RAM，Random Access Memory)存储器，也可为非不稳定的存储器(non-volatile memory)，例如磁盘存储器。存储器601用于存储一组可执行程序代码，处理器602与存储器601耦合。
进一步的，本申请实施例还提供了一种计算机可读存储介质，该计算机可读存储介质可以是设置于上述各实施例中的电子装置中，该计算机可读存储介质可以是前述实施例中的存储器601。该计算机可读存储介质上存储有计算机程序，该程序被处理器602执行时实现前述方法实施例中描述的机动目标的跟踪方法。
进一步的，该计算机可存储介质还可以是U盘、移动硬盘、只读存储器601(ROM，Read-Only Memory)、RAM、磁碟或者光盘等各种可以存储程序代码的介质。
在本申请所提供的几个实施例中，应该理解到，所揭露的装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述模块的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个模块或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或模块的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的模块可以是或者也可以不是物理上分开的，作为模块显示的部件可以是或者也可以不是物理模块，即可以 位于一个地方，或者也可以分布到多个网络模块上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。
另外，在本发明各个实施例中的各功能模块可以集成在一个处理模块中，也可以是各个模块单独物理存在，也可以两个或两个以上模块集成在一个模块中。上述集成的模块既可以采用硬件的形式实现，也可以采用软件功能模块的形式实现。
需要说明的是，对于前述的各方法实施例，为了简便描述，故将其都表述为一系列的动作组合，但是本领域技术人员应该知悉，本发明并不受所描述的动作顺序的限制，因为依据本发明，某些步骤可以采用其它顺序或者同时进行。其次，本领域技术人员也应该知悉，说明书中所描述的实施例均属于优选实施例，所涉及的动作和模块并不一定都是本发明所必须的。
在上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述的部分，可以参见其它实施例的相关描述。
以上为对本发明所提供的一种机动目标的跟踪方法及系统的描述，对于本领域的技术人员，依据本发明实施例的思想，在具体实施方式及应用范围上均会有改变之处，综上，本说明书内容不应理解为对本发明的限制。
工业实用性
解决了现有技术中对机动目标的准确估计上，估计结果与实际情况之间仍存在较大误差的技术问题。

Claims (10)

  1. 一种机动目标的跟踪方法，其特征在于，包括：
    基于T-S模糊语义模型估计模糊线性模型的状态预测值；
    根据所述状态预测值估计模糊线性模型的估计目标状态值；
    根据观测模型及所述估计目标状态值计算每个模糊模型的预测观测值；
    根据预测观测值组成的观测数据集计算每个模糊模型的模型模糊隶属度；
    根据离散动态系统的模糊语义模型计算目标的观测新息及航向角误差；
    将观测新息及航向角误差融入T-S模糊模型后更新前件参数；
    根据模型模糊隶属度计算前件参数的前件参数模糊隶属度；
    根据所述前件参数模糊隶属度组成的模糊集及离散动态系统计算每个线性模型的模型权值；
    根据模型权值计算目标的目标状态值，并根据目标状态值及模型权值计算目标的目标协方差；
    根据目标状态值及目标协方差估计机动目标的运动轨迹。
  2. 根据权利要求1所述的机动目标的跟踪方法，其特征在于，
    所述根据所述状态预测值估计模糊线性模型的估计目标状态值包括：
    引入最小二乘估计器；
    引入目标的目标速度和时间间隔作为最小二乘估计器内的遗忘因子；
    根据所述遗忘因子及最小二乘估计器建立修正的扩展遗忘因子最小二乘估计器，并根据修正的扩展遗忘因子最小二乘估计器及所述 状态预测值计算模糊线性模型的估计目标状态值。
  3. 根据权利要求1所述的机动目标的跟踪方法，其特征在于，
    所述根据预测观测值组成的观测数据集计算每个模糊模型的模型模糊隶属度包括：设定交叉熵；根据交叉熵设定模糊交叉熵；
    设定基于模糊交叉熵的核模糊C回归模型聚类的回归聚类函数；
    根据所述回归聚类函数及所述观测数据集计算每个模糊模型的模糊隶属度。
  4. 根据权利要求3所述的机动目标的跟踪方法，其特征在于，
    所述根据交叉熵设定模糊交叉熵包括：
    设定高斯函数为交叉熵的核函数；
    设定在小样本情况下，交叉熵的样本均值估计函数；
    根据所述样本均值估计函数及模糊信息处理理论，定义模糊交叉熵；
    根据模糊交叉熵辨识T-S模糊模型的前件参数。
  5. 根据权利要求4所述的机动目标的跟踪方法，其特征在于，
    所述设定基于模糊交叉熵的核模糊C回归模型聚类的回归聚类函数包括：
    根据观测数据集及模糊模型的输出设定核模糊C回归模型聚类的目标函数；
    设定目标函数的加权指数，并设定核空间距离函数；
    简化所述模糊交叉熵，并定义修正目标函数，且根据修正目标函数辨识T-S模糊模型的后件参数；
    将修正目标函数带入核空间距离函数，并根据所述前件参数及后件参数得到模糊隶属度函数，根据模糊隶属度函数计算每个模糊模型的模型模糊隶属度。
  6. 根据权利要求1所述的机动目标的跟踪方法及，其特征在于，
    所述将观测新息及航向角误差融入T-S模糊模型后更新前件参数包括：
    分别采用三个固定粒度的模糊集描述新息及航向角误差；
    使用高斯隶属函数表示所述固定粒度的模糊集；
    根据所述高斯隶属函数定义的前件参数更新T-S模糊模型，得到修改T-S模糊模型；
    根据修改T-S模糊模型及所述模型模糊隶属度更新前件参数。
  7. 根据权利要求1所述的机动目标的跟踪方法，其特征在于，
    所述T-S模糊模型的设定方法包括：
    设定离散非线性动态系统的非线性函数；
    使用模糊线性模型表示所述非线性函数；
    根据所述模糊线性模型得出全局模糊模型；
    设定钟型隶属函数为模糊隶属度函数，并根据所述钟型隶属度函数计算所述全局模糊模型内模型的模糊隶属度。
  8. 一种机动目标的跟踪系统，其特征在于，包括：
    状态预测值模块，用于基于T-S模糊语义模型估计模糊线性模型的状态预测值；
    目标状态值模块，用于根据所述状态预测值估计模糊线性模型的估计目标状态值；
    预测观测值模块，用于根据观测模型及所述估计目标状态值计算每个模糊模型的预测观测值；
    模型模糊隶属度模块，用于根据预测观测值组成的观测数据集计算每个模糊模型的模型模糊隶属度；
    参数模块，用于根据离散动态系统的模糊语义模型计算目标的观 测新息及航向角误差；
    更新前件参数模块，用于将观测新息及航向角误差融入T-S模糊模型后更新前件参数；
    前件参数模糊隶属度模块，用于根据模型模糊隶属度计算前件参数的前件参数模糊隶属度；
    模型权值模块，用于根据所述前件参数的模糊隶属度组成的模糊集及离散动态系统计算每个线性模型的模型权值；
    目标协方差模块，用于根据模型权值计算目标的目标状态值，并根据目标状态值及模型权值计算目标的目标协方差；
    运动轨迹估计模块，用于根据目标状态值及目标协方差估计机动目标的运动轨迹。
  9. 一种电子装置，包括：存储器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机程序，其特征在于，所述处理器执行所述计算机程序时，实现权利要求1至7中的任意一项所述方法。
  10. 一种计算机可读存储介质，其上存储有计算机程序，其特征在于，所述计算机程序被处理器执行时，实现权利要求1至7中的任意一项所述方法。
PCT/CN2018/118844 2018-12-03 2018-12-03 一种机动目标的跟踪方法及系统 WO2020113353A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2018/118844 WO2020113353A1 (zh) 2018-12-03 2018-12-03 一种机动目标的跟踪方法及系统

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2018/118844 WO2020113353A1 (zh) 2018-12-03 2018-12-03 一种机动目标的跟踪方法及系统

Publications (1)

Publication Number Publication Date
WO2020113353A1 true WO2020113353A1 (zh) 2020-06-11

Family

ID=70973424

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/118844 WO2020113353A1 (zh) 2018-12-03 2018-12-03 一种机动目标的跟踪方法及系统

Country Status (1)

Country Link
WO (1) WO2020113353A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111798491A (zh) * 2020-07-13 2020-10-20 哈尔滨工业大学 一种基于Elman神经网络的机动目标跟踪方法

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140343903A1 (en) * 2013-05-20 2014-11-20 Nec Corporation Factorial hidden markov models estimation device, method, and program
CN104483835A (zh) * 2014-11-06 2015-04-01 中国运载火箭技术研究院 一种基于t-s模糊模型的柔性航天器多目标综合控制方法
CN104808486A (zh) * 2015-02-13 2015-07-29 中国科学院自动化研究所 基于模糊ts模型的压电陶瓷执行器的预测控制方法和装置
CN104880708A (zh) * 2015-01-30 2015-09-02 西北工业大学 一种可变数目机动目标跟踪方法
CN105652250A (zh) * 2016-01-15 2016-06-08 西北工业大学 一种基于双层期望最大化的机动目标跟踪技术
CN105719312A (zh) * 2016-01-19 2016-06-29 深圳大学 基于序贯贝叶斯滤波的多目标跟踪方法及跟踪系统

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140343903A1 (en) * 2013-05-20 2014-11-20 Nec Corporation Factorial hidden markov models estimation device, method, and program
CN104483835A (zh) * 2014-11-06 2015-04-01 中国运载火箭技术研究院 一种基于t-s模糊模型的柔性航天器多目标综合控制方法
CN104880708A (zh) * 2015-01-30 2015-09-02 西北工业大学 一种可变数目机动目标跟踪方法
CN104808486A (zh) * 2015-02-13 2015-07-29 中国科学院自动化研究所 基于模糊ts模型的压电陶瓷执行器的预测控制方法和装置
CN105652250A (zh) * 2016-01-15 2016-06-08 西北工业大学 一种基于双层期望最大化的机动目标跟踪技术
CN105719312A (zh) * 2016-01-19 2016-06-29 深圳大学 基于序贯贝叶斯滤波的多目标跟踪方法及跟踪系统

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111798491A (zh) * 2020-07-13 2020-10-20 哈尔滨工业大学 一种基于Elman神经网络的机动目标跟踪方法

Similar Documents

Publication Publication Date Title
CN109990786B (zh) 机动目标跟踪方法及装置
CN109325128B (zh) 一种机动目标的跟踪方法及系统
CN110347971B (zh) 基于tsk模糊模型的粒子滤波方法、装置及存储介质
CN109558594B (zh) 交互式t-s模糊语义模型估计方法、系统和计算机可读存储介质
CN110111367B (zh) 用于目标跟踪的模型粒子滤波方法、装置、设备及存储介质
US20160161606A1 (en) Variational track management
CN111983927A (zh) 一种新型的最大协熵椭球集员滤波方法
CN111291471A (zh) 一种基于l1正则无迹变换的约束多模型滤波方法
Xu et al. Fixed-point iteration Gaussian sum filtering estimator with unknown time-varying non-Gaussian measurement noise
Yun et al. Robust Kalman filter with fading factor under state transition model mismatch and outliers interference
CN113452349B (zh) 一种基于贝叶斯序贯重要性积分的卡尔曼滤波方法
CN114445449A (zh) 基于变结构t-s模糊语义模型的目标跟踪方法及相关装置
WO2020113353A1 (zh) 一种机动目标的跟踪方法及系统
CN115204212A (zh) 一种基于stm-pmbm滤波算法的多目标跟踪方法
CN117057258B (zh) 基于权重分配相关系数的黑启动过电压预测方法及系统
CN112305915B (zh) Tsk迭代回归模型的标签多伯努利多目标跟踪方法及系统
Surya Maximum likelihood recursive state estimation: An incomplete-information based approach
Wang et al. A novel TS fuzzy particle filtering algorithm based on fuzzy C-regression clustering
Du et al. A novel locally regularized automatic construction method for RBF neural models
Cao et al. Generalized moving horizon estimation for nonlinear systems with robustness to measurement outliers
Wang et al. An adaptive outlier-robust Kalman filter based on sliding window and Pearson type VII distribution modeling
WO2020113354A1 (zh) 交互式t-s模糊语义模型估计方法、系统和计算机可读存储介质
Fu et al. Observation bootstrapping CDK-GMPHD filter based on consensus fusion strategy
Li et al. Interacting TS fuzzy semantic model estimation for maneuvering target tracking
CN112379350A (zh) 智能车辆毫米波雷达多目标跟踪方法、装置及设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18942197

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 17/09/2021)

122 Ep: pct application non-entry in european phase

Ref document number: 18942197

Country of ref document: EP

Kind code of ref document: A1