CN109190647B - An Active and Passive Data Fusion Method - Google Patents

Abstract

Translated fromChinese

本发明提供了一种有源无源数据融合方法，通过模糊聚类的方法，在时间维度和空间维度上根据有源数据和无源数据之间的相似度，在时间维上数据对准，空间维上模糊聚类，进行数据子集划分和关联隶属度计算。与现有技术相比，灵活性好，相较于现有方法中融合结果的好坏取决于角度差阈值门限的选取，不再需要设置专门的融合门限，使用更加灵活化；融合误差低，大大提高了融合可靠性，降低了融合误差；适应性好，可适用于所有的有源/无源数据融合系统，对于多种传感器的数据融合系统也可适用，具有很好的系统适用性。

The invention provides an active and passive data fusion method. Through the fuzzy clustering method, according to the similarity between the active data and the passive data in the time dimension and the space dimension, the data is aligned in the time dimension, Fuzzy clustering in spatial dimension, data subset division and association membership calculation. Compared with the existing technology, the flexibility is good. Compared with the existing method, the quality of the fusion result depends on the selection of the angle difference threshold. It is no longer necessary to set a special fusion threshold, and the use is more flexible; the fusion error is low, The fusion reliability is greatly improved and the fusion error is reduced; the adaptability is good, it can be applied to all active/passive data fusion systems, and it can also be applied to data fusion systems of various sensors, with good system applicability.

Description

Active and passive data fusion method

Technical Field

The invention relates to an active passive data fusion method, and relates to the field of radar/electronic warfare data fusion.

Background

In recent years, along with the improvement of weaponry and the improvement of informatization combat capability, a system integrating radar and electronic warfare becomes a development trend, so that a fusion technology based on active data and passive data becomes a research focus. The essence of the data fusion technique is that the information of different sensors is fused together through a fusion algorithm, so that more accurate information is obtained than that of a single sensor. By the data fusion processing technology, the space-time coverage range of the system can be expanded, the stability and the reliability of the system are enhanced, and the tracking capability and the recognition capability of the target are enhanced.

However, due to the reasons of mismatching of sensing ranges, mismatching of sensing dimensions, missing of information acquisition and the like between the active sensor and the passive sensor, an effective method suitable for data fusion of the active sensor and the passive sensor does not exist at present. Generally, an active sensor can provide more accurate target position information, while a passive sensor can provide rich radiation source attribute information, but the radiation source provided by the active sensor has weaker position capability, can only provide an arrival angle and has poorer accuracy. Most of the existing data fusion means are fusion based on position information, fusion judgment is carried out according to a set angle difference threshold value by comparing active and passive angle measurement values, the fusion result is limited by the value of the angle difference threshold value, the reliability is low, the error is large, and the application range of the technology is greatly limited.

Disclosure of Invention

The invention aims to solve the technical problem of providing an active and passive data fusion method, which can overcome the defects of low reliability and large error of a data fusion result caused by the limitation of the existing method to the setting of an angle difference threshold value, reduce the fusion error of active data and passive data and provide the fusion reliability.

The technical scheme adopted by the invention is as follows:

a fuzzy clustering method is used for carrying out data subset division and association membership calculation on time dimension and space dimension according to similarity between active data and passive data.

The specific method comprises the following steps of,

time fusion (1) is carried out on active data and passive data, and data integration is carried out on the active data and the passive data which are acquired in different observation times and scanning periods on a time dimension, wherein the data integration comprises time slice division and time alignment; the time alignment means that the active data and the passive data measurement time are uniformly sequenced in a selected time slice to form active measurement data and passive measurement data with uniformly sequenced measurement time;

acquiring a membership relation table (3), calculating the spatial statistical distance between active data and passive data corresponding to the same measuring time according to the corresponding measuring time based on a fuzzy clustering method, and constructing a fuzzy matrix to acquire the membership relation table;

and (4) performing association fusion on the active data and the passive data according to the membership relation table.

The method also comprises the steps of completing the active data (2), after time fusion, inquiring whether the moment has the measured active data or not according to the uniformly ordered measuring moments, if so, acquiring a membership table based on a fuzzy clustering method, and if not, calculating the active data at the moment.

The specific method for calculating the active data is to complement the active data corresponding to all the passive data measurement moments by means of interpolation or extrapolation according to the fused time sequence.

The time slice division is determined by the motion attribute of specific time, and the faster the target speed is, the shorter the divided time slice is.

The specific method for obtaining the membership table comprises the following steps,

obtaining a distance fuzzy subset of the active data and all the passive data corresponding to the active data at the same measuring moment according to all the passive data corresponding to the active data at the same measuring moment; the distance in the distance fuzzy subset is determined according to the distance square error of the space positions of the active data and the passive data;

selecting at least two membership functions for measurement according to the correlation degree of each distance between the active data and the passive data;

substituting each distance value in the distance fuzzy subset into the membership function to obtain a three-dimensional fuzzy matrix;

and weighting each membership degree corresponding to the same distance value in the three-dimensional fuzzy matrix, multiplying each membership degree by the corresponding weight value, and adding to obtain the membership relation of the distance values, thereby obtaining a membership relation table of active data and passive data.

The method also comprises the steps of refining the data of the fused result by utilizing the redundancy and complementarity between the active data and the passive data, counting the correlation between the fused results within a period of time, and giving a corresponding weight value, wherein the larger the correlation is, the larger the weight value is, and otherwise, the weight value is reduced.

And according to the correlation degree of each distance between the active data and the passive data, the membership function comprises four membership functions of good membership function, medium membership function and fair membership function.

The weights of four membership degrees of good, medium and fair which correspond to the same distance value in the three-dimensional fuzzy matrix are respectively 0.4, 0.3, 0.2 and 0.1.

Compared with the prior art, the invention has the beneficial effects that:

compared with the prior art that the quality of the fusion result depends on the selection of the threshold value threshold of the angle difference, the method adopts a space-time dimension fuzzy clustering based method, judges the fusion relation between the active data and the passive data through the membership degree relation table, does not need to set a special fusion threshold any more, and is more flexible to use.

The invention adopts fuzzy clustering method to obtain the membership degree association table of the active data and the passive data, associates according to the membership degree association table, gives certain weight to the association result, and updates the weight according to the long-time fusion result, thereby greatly improving the fusion reliability and reducing the fusion error.

The method is suitable for all active/passive data fusion systems, is also suitable for data fusion systems of various sensors and has good system applicability.

Drawings

Fig. 1 is a schematic flow chart of an active-passive fusion method according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

Any feature disclosed in this specification (including any accompanying drawings) may be replaced by alternative features serving equivalent or similar purposes, unless expressly stated otherwise. That is, unless expressly stated otherwise, each feature is only an example of a generic series of equivalent or similar features.

A fuzzy clustering method is used for carrying out data subset division and association membership calculation on time dimension and space dimension according to similarity between active data and passive data.

Compared with the prior art, the method has the advantages that a special angle difference threshold value does not need to be set, the fusion result has higher reliability, and the fusion error is greatly reduced.

The specific method comprises the following steps of,

as shown in fig. 1, time fusion is performed on active data and passive data (1), and data integration is performed on the active data and the passive data acquired in different observation times and scanning periods in a time dimension, including time slice division and time alignment; the time alignment means that the active data and the passive data measurement time are uniformly sequenced in a selected time slice to form active measurement data and passive measurement data with uniformly sequenced measurement time;

acquiring a membership relation table (3), calculating the spatial statistical distance between active data and passive data corresponding to the same measuring time according to the corresponding measuring time based on a fuzzy clustering method, and constructing a fuzzy matrix to acquire the membership relation table;

and (4) performing association fusion on the active data and the passive data according to the membership relation table.

The method also comprises the steps of completing the active data (2), after time fusion, inquiring whether the moment has the measured active data or not according to the uniformly ordered measuring moments, if so, acquiring a membership table based on a fuzzy clustering method, and if not, calculating the active data at the moment.

As an embodiment, the specific method for estimating the active data is to complement the active data corresponding to all the passive data measurement times by interpolation or extrapolation according to the fused time series.

Since the data rate of the passive data acquisition is higher than that of the active data, all the active data corresponding to the passive data measurement time need to be calculated for the convenience of subsequent fusion. During specific operation, according to the fused measurement time of unified sequencing, active data are screened and supplemented according to the passive measurement time: if the active data exists at the corresponding passive measurement time, continuing the subsequent operation, otherwise, completing the active data corresponding to the passive measurement time according to an interpolation or extrapolation mode.

And when no active measurement data exists at the corresponding passive measurement moment, performing active data completion by adopting an interpolation or extrapolation method. And during interpolation or extrapolation, a method for constructing an approximation function by adopting curve fitting is adopted, wherein the approximation function is set as y ═ f (t), t is the passive data measurement time, and y is the active data corresponding to the passive data measurement time. The concrete formula of the approximation function is as follows:

wherein, t₀、t₁、t₂Time instant, y, at which there is active measurement data for the time instant close to the passive data measurement instant₀＝f(t₀)、y₁＝f(t₁)、y₂＝f(t₂) Are each t₀、t₁、t₂Corresponding active metrology data, i.e. t₀、t₁、t₂、y₀、y₁、y₂Are all known quantities. Using t₀、t₁、t₂Time value and its corresponding function value y₀、y₁、y₂The second order polynomial coefficient of the approximation function y ═ f (t) can be determined and substituted into the passive measurement time value t to be interpolated or extrapolated, so that the active data corresponding to the passive measurement time can be further calculated.

The time slice division is determined by the motion attribute of specific time, the faster the target speed is, the shorter the divided time slice is, for example, the still/low speed target can select the fusion time slice to be minute, and the high speed target can select the fusion time slice to be second.

The specific method for obtaining the membership table comprises the following steps,

obtaining a distance fuzzy subset of the active data and all the passive data corresponding to the active data at the same measuring moment according to all the passive data corresponding to the active data at the same measuring moment; the distance in the distance fuzzy subset is determined according to the distance square error of the space positions of the active data and the passive data;

selecting at least two membership functions for measurement according to the correlation degree of each distance between the active data and the passive data;

substituting each distance value in the distance fuzzy subset into the membership function to obtain a three-dimensional fuzzy matrix;

and weighting each membership degree corresponding to the same distance value in the three-dimensional fuzzy matrix, multiplying each membership degree by the corresponding weight value, and adding to obtain the membership relation of the distance values, thereby obtaining a membership relation table of active data and passive data.

The method also comprises the steps of refining the data of the fused result by utilizing the redundancy and complementarity between the active data and the passive data, counting the correlation between the fused results within a period of time, and giving a corresponding weight value, wherein the larger the correlation is, the larger the weight value is, and otherwise, the weight value is reduced. And reducing the influence caused by error association according to the long-time fusion result and the weight value result, thereby obtaining a more accurate active data and passive data fusion result.

If the active data corresponding to the measurement time is P_i(i-1, …, N) corresponding to passive data being Q_j(j-1, …, M), one active data P is taken_iIt forms a finite fuzzy subset R ═ with respect to the statistical distance R (R) for all passive data at that moment_i1,r_i2,…,r_iM). The statistical distance r is determined according to the distance square difference of the space positions of the active data and the passive data; i is active data P_iThe number of a certain active data; j is a certain passive data Q in the passive data_jThe number of (2).

As a specific embodiment, the degree of correlation for each distance between active data and passive dataSelecting four membership functions u of good, medium, fair and poor₁、u₂、u₃、u₄The metric is performed and the membership function is expressed as follows.

Wherein:

in the above formula, w1, w2 and w3 are adjustment parameters, and are specifically determined according to the calculation range of the statistical distance r between the active data and the passive data, and four regions [0, w1], [ w1, w2], [ w2, w3], [ w3 and infinity ] of the statistical distance r divided by the adjustment parameters w1, w2 and w3 respectively correspond to four association degrees of good, medium, possible and poor distance association.

Substituting each distance value in the fuzzy subset R into the membership function u₁、u₂、u₃、u₄And obtaining a three-dimensional fuzzy matrix with one dimension of NxMx4.

For the same statistical distance r in the fuzzy matrix_ijThe corresponding 4 membership degrees are weighted according to good, medium, right and difference, and are sequentially taken as 0.4, 0.3, 0.2 and 0.1, and the statistical distance r is obtained by multiplication_ijCorresponding membership c_ij＝0.4*u₁(r_ij)+0.3*u₂(r_ij)+0.2*u₃(r_ij)+0.1*u₄(r_ij) Obtaining the membership degree relation table C of the active data and the passive data (C ═ C)_ij)_N×MThe membership degree relation table is a two-dimensional matrix with dimension NxM, and reflects the correlation condition of active data and passive data.

Claims (8)

Translated fromChinese

1.一种有源无源数据融合方法，通过模糊聚类的方法，在时间维度和空间维度上根据有源数据和无源数据之间的相似度，在时间维上数据对准，空间维上模糊聚类，进行数据子集划分和关联隶属度计算；1. An active and passive data fusion method, through the method of fuzzy clustering, according to the similarity between active data and passive data in time dimension and space dimension, data alignment in time dimension, spatial dimension. On fuzzy clustering, data subset division and association membership calculation are performed;具体方法包括，Specific methods include,对有源数据和无源数据进行时间融合（1），对在不同的观测时间和扫描周期内获取的有源数据和无源数据在时间维度上进行数据整合，包括时间片划分和时间对准；其中，所述时间对准是指，在选择的时间片内，对有源数据和无源数据量测时刻进行统一排序，形成量测时刻统一排序的有源量测数据和无源量测数据；Time fusion of active data and passive data (1), data integration of active data and passive data acquired in different observation times and scanning periods in the time dimension, including time slice division and time alignment ; Wherein, the time alignment means that, within the selected time slice, the measurement times of the active data and the passive data are uniformly sorted to form the active measurement data and the passive measurement which are uniformly sorted at the measurement times. data;获取隶属关系表（3），基于模糊聚类方法，按照对应量测时刻，计算同一量测时刻对应的有源数据与无源数据的空间统计距离，并构造模糊矩阵，获取隶属度关系表；Obtain the membership table (3), based on the fuzzy clustering method, according to the corresponding measurement time, calculate the spatial statistical distance between the active data and the passive data corresponding to the same measurement time, and construct a fuzzy matrix to obtain the membership table;关联融合（4），根据隶属度关系表，对有源数据与无源数据进行关联融合。Associative fusion (4), according to the membership table, the active data and the passive data are associated and fused.2.根据权利要求1所述的有源无源数据融合方法，所述方法还包括，补全有源数据（2），时间融合后，按照统一排序的量测时刻，查询该时刻是否存在量测的有源数据，如果是，则基于模糊聚类方法获取隶属关系表，如果否，则推算该时刻的有源数据。2 . The active and passive data fusion method according to claim 1 , the method further comprises: complementing the active data (2), and after the time fusion, according to the uniformly sorted measurement moments, inquiring whether there is an amount of data at the moment. 3 . The measured active data, if it is, the membership table is obtained based on the fuzzy clustering method, if not, the active data at this moment is estimated.3.根据权利要求2所述的有源无源数据融合方法，推算有源数据的具体方法为，按照融合后的时间序列，通过内插或外推的方法，补足对应所有无源数据量测时刻的有源数据。3. The active and passive data fusion method according to claim 2, the specific method for calculating the active data is, according to the time series after fusion, through interpolation or extrapolation, make up the corresponding measurement of all passive data. Moment of active data.4.根据权利要求1所述的有源无源数据融合方法，所述时间片划分以具体时间的运动属性确定，目标速度越快，划分的时间片越短。4 . The active and passive data fusion method according to claim 1 , wherein the time slice division is determined by the motion attribute of a specific time, and the faster the target speed, the shorter the divided time slice. 5 .5.根据权利要求1所述的有源无源数据融合方法，获取隶属关系表的具体方法包括，5. the active and passive data fusion method according to claim 1, the concrete method that obtains the affiliation table comprises,根据同一量测时刻某一有源数据对应的所有无源数据，得到该时刻有源数据与其所对应的所有无源数据的距离模糊子集；所述距离模糊子集中的距离根据有源数据与无源数据的空间位置的距离平方差确定；According to all the passive data corresponding to a certain active data at the same measurement time, the distance fuzzy subset of the active data and all the corresponding passive data at this time is obtained; the distance in the distance fuzzy subset is based on the difference between the active data and the passive data. Determine the distance squared difference of the spatial location of the passive data;根据有源数据与无源数据之间各个距离的关联程度，选择至少两种隶属度函数进行度量；According to the correlation degree of each distance between active data and passive data, at least two membership functions are selected for measurement;将距离模糊子集中的各个距离值带入所述隶属度函数，获得一个三维模糊矩阵；Bring each distance value in the distance fuzzy subset into the membership function to obtain a three-dimensional fuzzy matrix;对所述三维模糊矩阵中同一距离值对应的各个隶属度取权值，各个隶属度与其对应的权值相乘后相加得到所述距离值的隶属关系，从而获得有源数据与无源数据的隶属度关系表。Weights are taken for each membership degree corresponding to the same distance value in the three-dimensional fuzzy matrix, and each membership degree is multiplied by its corresponding weight value and added to obtain the membership relationship of the distance value, thereby obtaining active data and passive data. membership table.6.根据权利要求1所述的有源无源数据融合方法，所述方法还包括，利用有源数据和无源数据之间的冗余性和互补性，进行融合后结果的数据精炼，并统计持续一段时间内融合结果之间的相关性，赋予相应的权值，相关性越大，权值越大，反之则减小。6. The active and passive data fusion method according to claim 1, the method further comprises, utilizing the redundancy and complementarity between the active data and the passive data, to carry out the data refining of the result after fusion, and The correlation between the fusion results for a period of time is counted, and the corresponding weight is assigned. The greater the correlation, the greater the weight, and vice versa.7.根据权利要求5所述的有源无源数据融合方法，根据有源数据与无源数据之间各个距离的关联程度，所述隶属度函数包括好、中、可和差四种隶属度函数。7. The active and passive data fusion method according to claim 5, according to the degree of association of each distance between the active data and the passive data, the membership function includes four membership degrees of good, medium, acceptable and poor function.8.根据权利要求5所述的有源无源数据融合方法，三维模糊矩阵中同一距离值对应的好、中、可和差四种隶属度的权值，分别为0.4、0.3、0.2和0.1。8. The active and passive data fusion method according to claim 5, the weights of the four membership degrees of good, medium, acceptable and poor corresponding to the same distance value in the three-dimensional fuzzy matrix are 0.4, 0.3, 0.2 and 0.1 respectively. .

Images