CN115600051B - Orbit maneuver intelligent detection method and device based on short-arc space-based optical observation - Google Patents

Abstract

The application belongs to the technical field of space situational awareness, and relates to an orbit maneuver intelligent detection method and device based on short-arc space-based optical observation. The method includes: obtaining multiple groups of historical observation arcs belonging to different space objects to obtain the optimal arc; determining the initial orbits of the observation arcs adjacent in time according to the optimal arc; using the initial orbit as the initial value to perform least squares iteration , to obtain the orbit improvement result; convert the orbit improvement result into the orbit element, and obtain the maneuvering characteristic parameters; label each maneuvering characteristic parameter, and establish a label model; obtain the current observation arc through the short-arc space-based optical observation, and obtain Current maneuvering label: According to the current maneuvering label, the current observation arc segment belonging to the same space object is divided into the observation arc segment before maneuvering and the observation arc segment after maneuvering, the maneuvering parameters are estimated and the intelligent detection of orbital maneuvering is completed. The application enables the mobile detection of space-based short-arc optical observation segments.

Description

Orbit maneuver intelligent detection method and device based on short-arc space-based optical observation

technical field

The present application relates to the technical field of space situational awareness, in particular to an orbital maneuver intelligent detection method and device based on short-arc space-based optical observation.

Background technique

With the continuous development of aerospace technology, the maneuvering of in-orbit satellites is becoming more and more frequent. The development and implementation of various activities of in-orbit satellites are based on satellite orbital maneuvering. It is especially important, and has important applications in satellite behavior intention recognition and abnormal event perception.

If the orbital abnormal maneuvering behavior of key space targets can be detected early, the approaching events or other threats that may be unfavorable to our space targets can be avoided as soon as possible. Therefore, the maneuver detection of space targets has gradually become a important topic.

However, this problem faces two difficulties: one is that for pulsed orbital maneuvers, since the orbital velocity of the space target changes to a large extent before and after the maneuvering, the direct determination of the precise orbit of the arc segment of the observation data usually leads to divergence and As a result, it is difficult to converge and the orbit determination fails. Second, because the arc of observation data obtained by space-based optical observations through a single observation spans a relatively short length of time, the length of time usually does not exceed two minutes, or even within one minute. This arc segment of observation data is called a short-arc observation segment. Due to the short duration of a single short-arc observation segment, it is difficult to guarantee the accuracy of space object orbit determination. The above two difficulties make it difficult to realize the orbital pulse maneuver detection of space-based short-arc optical observation data in the existing technology.

Contents of the invention

Based on this, it is necessary to address the above technical problems and provide an orbital maneuvering intelligent detection method and device based on short-arc space-based optical observations, which can perform maneuvering detection on segments of space-based short-arc optical observations while taking into account the computational efficiency of the algorithm.

An intelligent detection method for orbital maneuvers based on short-arc space-based optical observations, including:

Obtain multiple groups of historical observation arcs belonging to different space objects, and use a quadratic polynomial to fit and screen the historical observation arcs to obtain an optimal arc;

According to the preferred arc segment, the observation arc segments belonging to the same space object are arranged in chronological order, and the initial orbits of the observation arc segments adjacent in time are determined;

Using the initial orbit as an initial value to perform least square iterations, and performing orbit improvement on observation arcs that belong to the same space target and are adjacent in time, and obtain an orbit improvement result;

Converting the orbital improvement result into an orbital element, calculating the variation of the orbital element, the previous orbital element, and the latter orbital element in the semi-major axis, eccentricity and orbital inclination, respectively, to obtain maneuvering characteristic parameters;

Putting a maneuvering label on each maneuvering characteristic parameter, and training the neural network according to the maneuvering characteristic parameter and the maneuvering label, and establishing a label model;

Obtain the current observation arc through short-arc space-based optical observation; obtain the current maneuvering tag of the current observation arc according to the current observation arc and the tag model;

According to the current maneuver label, the current observation arcs belonging to the same space target are divided into pre-maneuvering observation arcs and post-maneuvering observation arcs; according to the pre-maneuvering observation arcs and post-maneuvering observation arcs, maneuver parameters are estimated and orbital maneuver intelligence is completed detection.

In one embodiment, multiple groups of historical observation arcs belonging to different space objects are obtained, and a quadratic polynomial is used to fit and screen the historical observation arcs to obtain a preferred arc, specifically:

Obtain multiple groups of historical observation arcs belonging to different space objects, the historical observation arcs include: right ascension and declination;

Adopt quadratic polynomial to carry out fitting to the right ascension and the declination of described historical observation arc section, obtain the fitting coefficient of right ascension and the fitting coefficient of declination;

According to the fitting coefficient of right ascension and the fitting coefficient of declination, the standard deviation of the historical observation arc is obtained;

The historically observed arc segments are screened according to the standard deviation to obtain an optimal arc segment.

In one embodiment, a quadratic polynomial is used to fit the right ascension and declination of the historical observation arc to obtain a fitting coefficient of right ascension and a fitting coefficient of declination, specifically:

和赤纬

对时间的函数可以表示分别为：The quadratic polynomials are used to fit the function expressions of right ascension and declination in each historical observation arc segment with respect to time, and the right ascension

and Declination

The function of time can be expressed as:

（1）

(1)

，

，

为赤经的拟合系数，

，

，

为赤纬的拟合系数，各拟合系数的初值取为：in,

,

,

is the fitting coefficient of right ascension,

,

,

is the fitting coefficient of declination, and the initial value of each fitting coefficient is taken as:

（2）

(2)

对

，

，

的偏导数分别为：because

right

,

,

The partial derivatives of are respectively:

（3）

(3)

，

，

初值的改进量

，

，

为：Therefore, the least squares method can be used to obtain the pair

,

,

Improvement of the initial value

,

,

for:

（4）

(4)

是

的矩阵，

为

的转置矩阵，上标﹣1表示矩阵的求逆运算，

是

维的向量，

为赤经的多项式预测值；in,

yes

matrix,

for

The transposed matrix of , the superscript -1 represents the inverse operation of the matrix,

yes

dimension vector,

is the polynomial predicted value of right ascension;

，

，

更新为：then will

,

,

updated to:

（5）

(5)

小于设定的阈值，得到赤经的拟合系数

，

，

；Repeat the process of formula (4) and formula (5) until

is less than the set threshold, and the fitting coefficient of right ascension is obtained

,

,

;

执行以上相同的操作步骤，得到赤纬的拟合系数

，

，

。Declination

Perform the same operation steps above to get the fitting coefficient of declination

,

,

.

In one embodiment, according to the fitting coefficient of right ascension and the fitting coefficient of declination, the standard deviation of the historical observation arc segment is obtained, specifically:

，其中

表示对应观测弧段的中间行序号，由此对于每一个历史观测弧段

都有一个对应的中间时刻数据点：Define the middle moment of a historical observation arc as

,in

Indicates the serial number of the middle row corresponding to the observation arc, so for each historical observation arc

Each has a corresponding intermediate time data point:

（6）

(6)

为中间时刻赤经，

为中间时刻赤纬，

为中间时刻赤经变化率，

为中间时刻赤纬变化率，

，

分别为中间时刻对应光学观测卫星的位置矢量与速度矢量，其计算式如下：in

is the right ascension at the middle moment,

is the declination at the intermediate moment,

is the change rate of right ascension at the middle time,

is the change rate of declination at the middle time,

,

are the position vector and velocity vector corresponding to the optical observation satellite at the intermediate time, respectively, and their calculation formulas are as follows:

（7）

(7)

，进而得到标准差：For each observation moment, calculate the fitting value of right ascension and declination at the corresponding moment, and make a difference between the real observation value of right ascension and declination at the corresponding moment and the fitting difference, and the residual error of right ascension and declination can be obtained

, and then get the standard deviation:

表示第

个残差，

为残差均值，

为数据点个数。In the formula,

Indicates the first

residuals,

is the residual mean,

is the number of data points.

In one embodiment, a maneuver label is marked on each maneuver characteristic parameter, and the neural network is trained according to the maneuver characteristic parameter and the maneuver label, and a label model is established, specifically:

Randomly divide the maneuvering feature parameters in proportion to obtain the training set and test set of the neural network;

Put a maneuver label on each maneuver characteristic parameter;

According to the training set, the test set and the motorized label, supervised training is performed on the neural network to establish a label model.

In one embodiment, a maneuver label is placed on each maneuver characteristic parameter, specifically:

According to the actual maneuvering situation of the corresponding space target, a maneuvering label is marked on each maneuvering characteristic parameter. If the orbital maneuvering occurs in the time interval between the two observation arcs corresponding to the orbit element when the maneuvering characteristic parameter is calculated, the maneuvering characteristic The parameter is a maneuver characteristic parameter with maneuver, and the maneuver label is set to 1; otherwise, the maneuver characteristic parameter is a maneuver characteristic parameter without maneuver, and the maneuver label is set to 0.

In one embodiment, the current observation arc is obtained through short-arc space-based optical observation; according to the current observation arc and the tag model, the current maneuvering tag of the current observation arc is obtained, specifically:

Obtain the current observation arc segment through short-arc space-based optical observation;

The current observation arc is fitted and screened by a quadratic polynomial, and the initial orbit of the current observation arc adjacent in time is determined, and the orbit is improved for the current observation arc belonging to the same space target and adjacent in time , and get the current maneuvering characteristic parameters;

Input the current maneuver characteristic parameters into the label model to obtain the current maneuver label of the current observation arc.

In one embodiment, according to the current maneuver label, the current observation arc segment belonging to the same space object is divided into an observation arc segment before maneuvering and an observation arc segment after maneuvering, specifically:

Sort the current maneuvering characteristic parameters belonging to the same space target according to the observation time of the first data point of the first observation arc, and detect the current maneuvering label of the current maneuvering characteristic parameters;

If the current maneuver tags are all 0, then the space object has not performed pulse orbit maneuvers during the entire observation time interval;

If there is a current maneuvering tag of 1, then the two observation arcs of the orbit elements when calculating the current maneuvering characteristic parameters corresponding to the current maneuvering tag are used as the dividing points, and the observation arcs to which the space object belongs are divided into The observation arc before the maneuver and the observation arc after the maneuver, and the time interval between the observation arc before the maneuver and the observation arc after the maneuver is defined as the time interval for predicting the application of the pulse maneuver.

In one embodiment, according to the observation arc before maneuvering and the observation arc after maneuvering, maneuver parameters are estimated and orbit maneuver intelligent detection is completed, specifically:

According to the observation arc before the maneuver and the observation arc after the maneuver, the least squares trajectory is iteratively improved, and the trajectory improvement result before the maneuver and the trajectory improvement result after the maneuver are obtained;

The trajectory improvement results before the maneuver and the trajectory improvement results after the maneuver are traversed through the track crossing forecast in the estimated time interval of the impulse maneuver application, and the maximum likelihood time of the impulse maneuver application is obtained;

According to the maximum likelihood moment, the magnitude of the pulse maneuver and the application direction of the pulse maneuver are estimated to complete the intelligent detection of the orbital maneuver.

An orbital maneuver intelligent detection method device based on short-arc space-based optical observation, including:

The obtaining module is used to obtain multiple groups of historical observation arcs belonging to different space objects, and adopts a quadratic polynomial to fit and screen the historical observation arcs to obtain a preferred arc;

The arrangement module is used to arrange the observation arcs belonging to the same space object in chronological order according to the preferred arcs, and determine the initial orbits of the observation arcs adjacent in time;

The iteration module is used to perform least squares iteration using the initial orbit as an initial value, and perform orbit improvement on observation arcs that belong to the same space target and are adjacent in time to obtain an orbit improvement result;

A calculation module, configured to convert the orbital improvement result into an orbital element, and calculate the changes in the semi-major axis, eccentricity, and orbital inclination of the orbital element, the previous orbital element, and the subsequent orbital element, respectively, Get the maneuvering characteristic parameters;

A modeling module is used to label each maneuver feature parameter, and train the neural network according to the maneuver feature parameter and the maneuver label to establish a label model;

The label module is used to obtain the current observation arc through short-arc space-based optical observation; obtain the current maneuvering label of the current observation arc according to the current observation arc and the label model;

The estimation module is used to divide the current observation arcs belonging to the same space target into pre-maneuvering observation arcs and post-maneuvering observation arcs according to the current maneuvering label; estimate maneuvering parameters according to the pre-maneuvering observation arcs and post-maneuvering observation arcs And complete the track maneuver intelligent detection.

The above-mentioned orbital maneuvering intelligent detection method based on short-arc space-based optical observation obtains multiple sets of historical observation arcs belonging to different space objects, and uses quadratic polynomials to fit and filter the historical observation arcs. The arcs are arranged in chronological order, and the initial orbits of the observation arcs adjacent in time are determined, and the least squares iteration is performed with the initial orbit as the initial value, and the orbits of the observation arcs that belong to the same space target and are adjacent in time are improved. Obtain the orbit improvement result, convert the orbit improvement result into the orbit element, calculate the variation of the orbit element, the previous orbit element and the latter orbit element in the semi-major axis, eccentricity and orbit inclination, and obtain the maneuvering Characteristic parameters; mark each maneuvering characteristic parameter with a maneuvering label, and train the neural network according to the maneuvering characteristic parameter and maneuvering label, and establish a label model; obtain the current observation arc segment through short-arc space-based optical observation; according to the current observation arc segment and label model to obtain the current maneuvering label of the current observation arc; according to the current maneuvering label, the current observation arc belonging to the same space object is divided into the pre-maneuvering observation arc and the post-maneuvering observation arc; according to the pre-maneuvering observation arc And observe the arc after maneuvering, estimate the maneuvering parameters and complete the intelligent detection of orbital maneuvering (the intelligent detection of orbital maneuvering includes: judging whether there is orbital maneuvering according to the maneuvering label, and estimating the maneuvering parameters). This application is applicable to the space target pulse orbit maneuver detection under space-based optical short-arc observation conditions. By extracting the key features of the space target orbit before and after the maneuver, the maneuver characteristic parameters are constructed, thereby realizing the effective training of the neural network, and ingeniously avoiding the In addition to the problem of maneuver detection threshold design, the calculation efficiency and calculation accuracy of pulse track maneuver detection are also taken into account.

Description of drawings

Fig. 1 is an application scenario diagram of an orbital maneuver intelligent detection method based on short-arc space-based optical observation in an embodiment;

Figure 2 is a schematic flow chart of an orbital maneuvering intelligent detection method based on short-arc space-based optical observations in one embodiment;

Fig. 3 is a structural block diagram of an orbital mobile intelligent detection device based on short-arc space-based optical observation in an embodiment.

Detailed ways

In order to make the purpose, technical solution and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present application, and are not intended to limit the present application. Based on the embodiments in this application, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the scope of protection of this application.

It should be noted that all directional indications (such as up, down, left, right, front, back...) in the embodiments of the present application are only used to explain the relationship between the components in a certain posture (as shown in the figure). Relative positional relationship, movement conditions, etc., if the specific posture changes, the directional indication will also change accordingly.

In addition, descriptions such as "first", "second" and so on in this application are only for description purposes, and should not be understood as indicating or implying their relative importance or implicitly indicating the number of indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include at least one of these features. In the description of the present application, "multiple groups" means at least two groups, such as two groups, three groups, etc., unless specifically defined otherwise.

In this application, unless otherwise clearly specified and limited, the terms "connection" and "fixation" should be interpreted in a broad sense, for example, "fixation" can be a fixed connection, a detachable connection, or an integral body; It can be a mechanical connection, an electrical connection, a physical connection or a wireless communication connection; it can be a direct connection or an indirect connection through an intermediary, and it can be an internal connection between two components or an interaction relationship between two components. Unless expressly defined otherwise. Those of ordinary skill in the art can understand the specific meanings of the above terms in this application according to specific situations.

In addition, the technical solutions of the various embodiments of the present application can be combined with each other, but it must be based on the realization of those skilled in the art. When the combination of technical solutions is contradictory or cannot be realized, it should be considered as a combination of technical solutions. Does not exist, nor is it within the scope of protection required by this application.

The method provided in this application can be applied to the application environment shown in FIG. 1 . Among them, the terminal 102 communicates with the server 104 through the network, the terminal 102 can include but not limited to various personal computers, notebook computers, smart phones, tablet computers and portable wearable devices, the server 104 can be various portal websites, work systems The corresponding server in the background, etc.

This application provides an orbital maneuver intelligent detection method based on short-arc space-based optical observation, as shown in Figure 2. In one embodiment, the method is applied to the terminal in Figure 1 as an example, including:

In step 201, multiple sets of historical observation arcs belonging to different space objects are acquired, and a quadratic polynomial is used to fit and filter the historical observation arcs to obtain an optimal arc.

specific:

Obtain multiple groups of historical observation arcs belonging to different space targets, the historical observation arcs include: right ascension and declination; use a quadratic polynomial to fit the right ascension and declination of the historical observation arcs to obtain declination The fitting coefficient of longitude and the fitting coefficient of declination; According to the fitting coefficient of right ascension and the fitting coefficient of declination, obtain the standard deviation of the historical observation arc; Screen to get the optimal arc segment.

more specific:

，

，

，其中

为空间目标数量，

为观测弧段的数量，

，

为第

个空间目标的第

个观测弧段的数据行数，下标

表示弧段中第

行数据，

为观测历元时刻，

为赤经，

为赤纬，

与

为分别为每行数据观测历元时刻对应的光学观测卫星的位置和速度矢量。Step 2011, it is known that the space-based observation equipment installed on the low-orbit optical observation satellite is used to conduct long-term optical observations, and after performing arc correlation matching, the obtained multiple groups of observation arcs belonging to different space objects (that is, historical observation arcs Each observation arc segment data includes several observation data points, and each observation data point is composed of the right ascension, declination, observation time and position and velocity information of the observed target relative to the low-orbit optical observation satellite. space-based goniometric data

,

,

,in

is the number of spatial objects,

is the number of observation arcs,

,

for the first

the first space object

The number of data rows of each observation arc segment, subscript

Indicates the arc segment

row data,

To observe the epoch time,

is right ascension,

is the declination,

and

are the position and velocity vectors of the optical observation satellite corresponding to the observation epoch time of each row of data, respectively.

Step 2012, using quadratic polynomials to respectively fit the function expressions of right ascension and declination with respect to time in each observation arc segment, so as to obtain the change rate information of right ascension and declination with time.

和赤纬

对时间的函数可以表示分别为：Set right ascension

and Declination

The function of time can be expressed as:

（1）

(1)

，

，

，

，

，

为多项式待定系数，

，

，

为赤经的拟合系数，

，

，

为赤纬的拟合系数，各待定系数的初值可以取为：That

,

,

,

,

,

are polynomial undetermined coefficients,

,

,

is the fitting coefficient of right ascension,

,

,

is the fitting coefficient of declination, and the initial value of each undetermined coefficient can be taken as:

（2）

(2)

对

，

，

的偏导数分别为：because

right

,

,

The partial derivatives of are respectively:

（3）

(3)

，

，

初值的改进量

，

，

为：Therefore, the least squares method can be used to obtain the pair

,

,

Improvement of the initial value

,

,

for:

（4）

(4)

是

的矩阵，

为

的转置矩阵，上标﹣1表示矩阵的求逆运算，

是

维的向量，

为赤经的多项式预测值：in

yes

matrix,

for

The transposed matrix of , the superscript -1 represents the inverse operation of the matrix,

yes

dimension vector,

is the polynomial predictor of right ascension:

，

，

更新为：then will

,

,

updated to:

（5）

(5)

小于设定的阈值即可，一般可取为

，最终可以得到拟合出的赤经的拟合系数

，

，

；将赤纬

执行以上相同的操作步骤可以得到相应的赤纬的拟合系数

，

，

。Repeat the process of formula (4) and formula (5) until

is less than the set threshold, generally it can be taken as

, and finally the fitting coefficient of the fitted right ascension can be obtained

,

,

; will declination

Perform the same operation steps above to get the fitting coefficient of the corresponding declination

,

,

.

，其中

表示对应观测弧段的中间行序号，由此对于每一个观测弧段

都有一个对应的中间时刻数据点：Step 2013, defining the intermediate moment of an observation arc as

,in

Indicates the serial number of the middle row corresponding to the observation arc, so for each observation arc

Each has a corresponding intermediate time data point:

（6）

(6)

为中间时刻赤经，

为中间时刻赤纬，

为中间时刻赤经变化率，

为中间时刻赤纬变化率，

，

分别为中间时刻对应光学观测卫星的位置矢量与速度矢量。其计算式如下：in

is the right ascension at the middle moment,

is the declination at the intermediate moment,

is the change rate of right ascension at the middle time,

is the change rate of declination at the middle time,

,

are the position vector and velocity vector of the optical observation satellite corresponding to the intermediate moment, respectively. Its calculation formula is as follows:

（7）

(7)

Step 2014. For each observation time, the fitting value of right ascension and declination at the corresponding time can be obtained through formula (1), and the real observation value of right ascension and declination at the corresponding time is made a difference with the fitting value, and right ascension and declination can be obtained residuals. According to the formula for calculating the population standard deviation:

（8）

(8)

，其中，

表示第

个残差，

为残差均值，

为数据点个数。若某数据点的残差大于

可以认定该点为坏点，应当将该点进行剔除，否则可能会影响后续轨道改进的精度甚至导致轨道改进不收敛。From this, the standard deviation of the fitted value of an arc segment and the residual of the actual observed value can be calculated

,in,

Indicates the first

residuals,

is the residual mean,

is the number of data points. If the residual of a data point is greater than

This point can be identified as a bad point and should be eliminated, otherwise it may affect the accuracy of subsequent track improvement or even lead to non-convergence of track improvement.

Step 202, according to the preferred arc segment, arrange the observation arc segments belonging to the same space object in chronological order, and determine the initial orbits of the observation arc segments adjacent in time.

specific:

，

，

，其中

为空间目标数量，

为观测弧段的数量。Step 2021, sort the observation arcs belonging to the same space object according to the order of the observation time of the first data point, and assume that the space-based angle measurement data arc obtained after sorting is

,

,

,in

is the number of spatial objects,

is the number of observation arcs.

Step 2022, determine an initial orbit for every two adjacent observation arcs after time sorting, if the time difference between the adjacent two arcs is too large and it is difficult to combine the two observation arcs to determine the initial orbit, also The result obtained by determining the initial orbit of one of the observation arcs can be used as the corresponding initial orbit.

It should be noted that for the observation data of angle measurement only, the initial orbit determination algorithm is a mature algorithm in the field of aerospace dynamics, for example, you can refer to the initial orbit determination method in the following literature: Liu Lin, Hu Songjie, Cao Jianfeng, Tang Jingshi. Spacecraft Orbit Determination Theory and Application [M]. Electronic Industry Press, 2015, Beijing.

，

，

，其中

为空间目标数量，

为观测弧段的数量，

，

为定轨历元时刻，一般选取为首个数据点的观测时刻或整个观测弧段的中间时刻，

与

为分别为定轨历元时刻对应的空间目标的位置和速度矢量。Suppose the obtained initial orbit result is

,

,

,in

is the number of spatial objects,

is the number of observation arcs,

,

is the orbit determination epoch time, generally selected as the observation time of the first data point or the middle time of the entire observation arc,

and

are the position and velocity vectors of the space object corresponding to the orbit determination epoch, respectively.

In step 203, the least squares iteration is performed with the initial orbit as an initial value, and the orbit improvement is performed on the observation arc segments that belong to the same space target and are adjacent in time to obtain an orbit improvement result.

specific:

Step 2031, perform least squares iterative orbit improvement between adjacent observation arcs sorted by time, the initial value of the iteration is the initial orbit determination result obtained in step 2022, and the least squares iterative orbit improvement algorithm is also an aerospace For mature algorithms in the field of dynamics, for example, refer to the orbit improvement methods in the following literature: Liu Lin, Hu Songjie, Cao Jianfeng, Tang Jingshi. Spacecraft Orbit Determination Theory and Application [M]. Electronic Industry Press, 2015, Beijing.

，

，

，其中

为空间目标数量，

为观测弧段的数量，

，

，

与

的定义与步骤2022中的

，

与

定义相同。Step 2032, store all the track improvement results obtained, and assume that the track improvement results obtained are

,

,

,in

is the number of spatial objects,

is the number of observation arcs,

,

,

and

The definition of and in step 2022

,

and

Same definition.

个观测弧段的空间目标，只会存在

条初轨确定结果与轨道改进结果。It should be noted that since the initial orbit determination and orbit improvement are carried out between adjacent observation arcs, for a certain existence

The space target of observation arc segment will only exist

The results of initial orbit determination and orbit improvement.

Step 204, converting the orbit improvement result into orbital elements, calculating the changes in the semi-major axis, eccentricity, and orbital inclination of the orbital elements, the previous orbital elements, and the subsequent orbital elements, respectively, to obtain maneuvering Characteristic Parameters.

specific:

Convert the orbital improvement result into orbital elements, and for every three groups of orbital elements that belong to the same space target and are adjacent in time, the first group and the second group, the second group and the third group (the orbit The number of roots is the second group, the number of the previous orbit is the first group, and the number of the latter orbit is the third group) and the semi-major axis, eccentricity and orbital inclination are differenced to obtain the semi-major axis and eccentricity of the two groups respectively and the variation of orbital inclination, and record the variation values of these two groups of semi-major axis, eccentricity and orbital inclination as a maneuver characteristic parameter.

more specific:

，

，

，其中

为空间目标数量，

为观测弧段的数量，

，

分别对应轨道根数中的半长轴、偏心率、轨道倾角、升交点赤经、近地点辐角与真近点角。由空间目标的位置速度计算轨道根数的计算式为：Step 2041, convert the obtained orbital improvement result into an orbital element, and assume that the obtained orbital element result is

,

,

,in

is the number of spatial objects,

is the number of observation arcs,

,

Corresponding to the semi-major axis, eccentricity, orbital inclination, right ascension of ascending node, argument of perigee and true anomaly in orbital elements respectively. The formula for calculating orbital elements from the position and velocity of space objects is:

（9）

(9)

为地心引力常量，

为升交点幅角。in

is the gravitational constant,

is the argument of ascending node.

，

，

，其中为空间目标数量，为观测弧段的数量。机动特征参数的具体定义为

，其中

分别表示半长轴、偏心率、轨道倾角的变化量，其计算式如下所示：Step 2042, taking the difference between the first group and the second group, the second group and the third group's semi-major axis, eccentricity and orbital inclination angle in every three groups of orbital radicals that are adjacent in time to obtain two and a half groups For the variation of the major axis, eccentricity and orbital inclination, the absolute value of the two sets of semi-major axis, eccentricity and orbital inclination are recorded as a maneuvering characteristic parameter and stored, and the result of the obtained maneuvering characteristic parameter is assumed to be

,

,

, where is the number of space targets, and is the number of observation arcs. The specific definition of the maneuvering characteristic parameters is

,in

Respectively represent the variation of the semi-major axis, eccentricity, and orbital inclination, and their calculation formulas are as follows:

（10）

(10)

个观测弧段的空间目标，由于经步骤201至步骤203以及步骤2041会计算得到

条轨道根数，且每相邻的三组轨道根数才能计算得到一条机动特征参数，因此对于该空间目标，只会存在

条机动特征参数。It should be noted that for an existing

The space target of the observation arc segment will be calculated by step 201 to step 203 and step 2041

The number of orbital elements, and every three adjacent sets of orbital elements can be calculated to obtain a maneuvering characteristic parameter, so for this space target, there will only be

Bar maneuver characteristic parameters.

In step 205, a maneuver label is attached to each maneuver feature parameter, and a neural network is trained according to the maneuver feature parameter and the maneuver label to establish a label model.

specific:

Randomly divide the maneuvering characteristic parameters in proportion to obtain the training set and test set of the neural network; according to the actual maneuvering situation of the corresponding space target, put a maneuvering label on each maneuvering characteristic parameter. In the time interval between two observation arcs of the orbital elements (that is, the second group of orbital elements), the maneuvering characteristic parameter is a maneuvering characteristic parameter, and the maneuvering label is set to 1; otherwise, the maneuvering characteristic The parameter is a maneuver characteristic parameter without maneuver, and the maneuver label is set to 0; according to the training set, the test set and the maneuver label, supervised training is carried out to the neural network, and a label model is established.

more specific:

Step 2051. According to the space-based optical observation data of known space objects (generally not less than 1000 observation arcs, which must also contain space objects that have undergone pulse orbit maneuvers), follow the first four steps (step 201-step 204) ) to process and finally obtain multiple maneuver characteristic parameters.

Step 2052, according to the actual maneuvering situation of the corresponding space object, put a maneuvering label on each maneuvering characteristic parameter, if the orbital maneuvering occurs in the time interval between two observation arcs corresponding to the second group of orbital elements when calculating the maneuvering characteristic parameter , then the maneuver characteristic parameter is considered to be a maneuver characteristic parameter with maneuver, and the maneuver label is set to 1; otherwise, the maneuver characteristic parameter is considered to be a maneuver characteristic parameter without maneuver, and the maneuver label is set to 0.

It should be noted that if the moment of applying the pulse maneuver of a certain space object is located in the time interval between the first and second observation arc of the space object or the time interval or observation arc between the penultimate first and second observation arc If the number of segments is less than 6, the observation data of such space objects should be discarded because the maneuvering characteristic parameters formed by such space objects are difficult to provide effective learning samples for the training of the neural network.

Step 2053: Randomly divide the large number of obtained maneuvering feature parameters in proportion to obtain the training set T and test set D required for neural network training, which can be divided by using a division method such as the hold-out method or the cross-validation method.

Step 2054, conduct neural network training, use the training set T and test set D generated in step 2053 to conduct supervised training with a feedforward neural network model (FNN, Forward Neural Network); the network model recommends using cascaded feedforward neural networks Network model (CFNN, Cascade Forward Neural Network), the training goal of the neural network is to be able to output the maneuver label (0 or 1) corresponding to each maneuver characteristic parameter according to the input maneuver characteristic parameters, that is, the corresponding maneuver characteristic parameter calculation. Whether impulsive maneuvering is applied in the time interval between two observation arcs of the second group of orbital elements.

There are a variety of feed-forward neural network models that can be directly called in the Neural Network Tool that comes with Matlab. For detailed information about the cascaded feed-forward neural network model, you can read the following literature: De JesusO, Hagan M T. Backpropagation Algorithms for a Broad Class of Dynamic Networks[J]. IEEE Transactions on Neural Networks, 2007, 18(1):14-27.

In step 206, the current observation arc is obtained through short-arc space-based optical observation; and the current maneuvering tag of the current observation arc is obtained according to the current observation arc and the tag model.

specific:

Through short-arc space-based optical observations, the current observation arc is obtained; the current observation arc is fitted and screened by a quadratic polynomial, and the initial orbit of the current observation arc adjacent in time is determined. The orbit of the target and temporally adjacent current observation arc is improved, and the current maneuver characteristic parameters are obtained; the current maneuver characteristic parameters are input into the label model, and the current maneuver label of the current observation arc is obtained.

more specific:

Input the real maneuvering feature parameters to be detected for maneuvering obtained after the real observation data are processed in steps 201 to 204 into the neural network trained in step 205, for each input real maneuvering feature parameter, the trained neural network The corresponding maneuvering labels will be output, and each input maneuvering feature parameter will be corresponded with the corresponding output maneuvering labels one by one and recorded and saved.

Step 207, according to the current maneuver label, divide the current observation arc segment belonging to the same space object into pre-maneuvering observation arc segment and post-maneuvering observation arc segment; according to the pre-maneuvering observation arc segment and post-maneuvering observation arc segment, estimate maneuver parameters and complete Track maneuver intelligent detection.

specific:

Sort the current maneuvering characteristic parameters belonging to the same space target according to the observation time of the first data point of the first observation arc, and detect the current maneuvering label of the current maneuvering characteristic parameter; if the current maneuvering label is all 0, the space The target does not perform pulse orbit maneuvers during the entire observation time interval; if there is a current maneuver tag with a value of 1, the two observation arcs of the orbital elements corresponding to the current maneuver tag when calculating the current maneuver characteristic parameters are used as the dividing point , the observation arc to which the space target belongs is divided into the observation arc before the maneuver and the observation arc after the maneuver, and the time interval between the observation arc before the maneuver and the observation arc after the maneuver is defined as the estimated pulse maneuver application time interval.

According to the observation arc before the maneuver and the observation arc after the maneuver, the least squares orbit improvement is carried out iteratively, and the orbit improvement results before the maneuver and the orbit improvement results after the maneuver are obtained; the orbit improvement results before the maneuver and the orbit improvement results after the maneuver are estimated The time interval of maneuver application is traversed through the track crossing forecast to obtain the maximum likelihood moment of impulse maneuver application; according to the maximum likelihood moment, the magnitude of impulse maneuver and the direction of impulse maneuver application are estimated to complete the intelligent detection of orbit maneuver.

more specific:

Step 2071, sort the real maneuvering characteristic parameters belonging to the same space target according to the observation time of the first data point of the first observation arc, and then sequentially detect the maneuvering labels of each maneuvering characteristic parameter, if the maneuvering labels are all 0, Then it can be considered that the target has not performed pulse orbital maneuvers in the entire observation time interval; if there is a maneuvering label of 1, the two observation arcs corresponding to the second group of orbital elements in the calculation of the maneuvering characteristic parameters are used as the dividing point , that is, according to the maneuver label of the real maneuver characteristic parameter, the observation arc segment to which the space object belongs is divided into the observation arc segment before the maneuver and the observation arc segment after the maneuver, and the time interval between these two observation arc segments is defined as the estimated Pulse maneuver application time interval.

For example, if a space object A has 8 observation arcs in the observation time interval, 5 maneuver characteristic parameters are obtained after calculation from step 201 to step 204, and it is detected that the maneuver label corresponding to the third maneuver characteristic parameter is 1, The second set of orbit elements corresponding to the calculation of the third maneuver characteristic parameter is obtained by improving the orbits of the third and fourth observation arcs, so the third and fourth observation arcs are taken as the dividing point, and the The three arcs are divided into observation arcs before maneuvering, the fourth to eighth arcs are divided into observation arcs after maneuvering, and the time interval between the third and fourth observation arcs is the estimated pulse maneuver application time interval.

，

，

，其中

为可能存在脉冲轨道机动的空间目标数量，

表示机动前，

表示机动后，

，

，

与

的定义与步骤2022中的定义相同；关于最小二乘轨道迭代改进算法的相关信息详见步骤2031。Step 2072: Combine all observation arcs before maneuvering and all observation arcs after maneuvering of the same space object to perform least squares iterative orbit improvement. Therefore, for a space target that is considered to have impulsive orbital maneuvering, the pre-maneuvering and maneuvering The improvement results of the last two orbits, and the obtained orbit improvement results are as follows:

,

,

,in

is the number of space targets that may have impulsive orbital maneuvers,

Indicates before maneuvering,

Indicates that after maneuvering,

,

,

and

The definition of is the same as that in step 2022; for the relevant information about the iterative improvement algorithm of the least squares trajectory, see step 2031.

，其中

，

为可能存在脉冲轨道机动的空间目标数量，预估脉冲机动施加时间区间的定义与步骤2071中定义一致。轨道交叉预报具体是指将两个空间目标轨道分别进行向前和向后轨道预报以预报同一历元时刻的操作。此处脉冲机动施加的极大似然时刻

定义为使得机动前轨道与机动后轨道在三维空间中位置距离最小的时刻。此定义可以进行改进与拓展，如若机动前轨道与机动后轨道的轨道偏差信息已知，脉冲机动施加的极大似然时刻

还可定义为使得机动前轨道与机动后轨道马氏距离最小的时刻。Step 2073: Traversing the trajectory improvement results before maneuvering and post-maneuvering trajectory calculated in step 2072 through orbit crossing prediction within the estimated pulse maneuver application time interval, to find the maximum likelihood moment of pulse maneuver application

,in

,

For the number of space targets that may have impulsive orbital maneuvers, the definition of the estimated time interval for applying impulsive maneuvers is consistent with the definition in step 2071 . Orbit crossing prediction specifically refers to the operation of predicting the orbits of two space objects forward and backward respectively to predict the same epoch time. Here the maximum likelihood moment of impulsive maneuvering is applied

It is defined as the moment when the positional distance between the trajectory before the maneuver and the trajectory after the maneuver is the smallest in three-dimensional space. This definition can be improved and expanded. If the orbit deviation information of the orbit before the maneuver and the orbit after the maneuver is known, the maximum likelihood moment imposed by the impulsive maneuver

It can also be defined as the moment when the Mahalanobis distance between the orbit before the maneuver and the orbit after the maneuver is the smallest.

。Step 2074: Predict the trajectory improvement results before the maneuver and the trajectory improvement results after the maneuver respectively to the maximum likelihood moment of the impulsive maneuver, and conduct the impulsive maneuver by comparing and analyzing the trajectory improvement results before the maneuver and the trajectory improvement results after the maneuver The estimation of other maneuvering parameters such as size, direction of pulse maneuvering application, that is: the trajectory improvement results before maneuvering and the orbital improvement results after maneuvering are respectively carried out backward trajectory prediction and forward trajectory prediction, and the prediction is obtained in step 2073. maximum likelihood moment

.

与

，施加的脉冲机动冲量可以通过下式进行估算：Assume that the orbit forecast results before maneuvering and after maneuvering are respectively:

and

, the applied pulse motor impulse can be estimated by the following formula:

（11）

(11)

表示脉冲机动冲量的估计值。in

Indicates the estimated value of the impulse motor momentum.

时，该检测结果有很大概率是误检，应当认为该次脉冲机动不存在。Step 2075. In actual observation, due to disturbance factors such as observation errors caused by certain reasons, a small number of false detections may occur. Therefore, in order to improve the detection accuracy of orbital maneuvers, it is also necessary to correct Filter the test results. According to practical experience, when

, the detection result has a high probability of false detection, and it should be considered that the impulse maneuver does not exist.

The above-mentioned orbital maneuvering intelligent detection method based on short-arc space-based optical observation obtains multiple sets of historical observation arcs belonging to different space objects, and uses quadratic polynomials to fit and filter the historical observation arcs. The arcs are arranged in chronological order, and the initial orbits of the observation arcs adjacent in time are determined, and the least squares iteration is performed with the initial orbit as the initial value, and the orbits of the observation arcs that belong to the same space target and are adjacent in time are improved. Obtain the orbit improvement result, convert the orbit improvement result into the orbit element, calculate the variation of the orbit element, the previous orbit element and the latter orbit element in the semi-major axis, eccentricity and orbit inclination, and obtain the maneuvering Characteristic parameters; mark each maneuvering characteristic parameter with a maneuvering label, and train the neural network according to the maneuvering characteristic parameter and maneuvering label, and establish a label model; obtain the current observation arc segment through short-arc space-based optical observation; according to the current observation arc segment and label model to obtain the current maneuvering label of the current observation arc; according to the current maneuvering label, the current observation arc belonging to the same space object is divided into the pre-maneuvering observation arc and the post-maneuvering observation arc; according to the pre-maneuvering observation arc And observe the arc section after maneuvering, estimate maneuvering parameters and complete orbital maneuvering intelligent detection.

This application is applicable to space target pulse track maneuver detection under space-based optical short-arc observation conditions. The neural network trained by constructing the maneuver characteristic parameters can detect which two observation arcs the maneuver occurs between, so as to obtain The observation arcs are divided into observation arcs before maneuvering and observation arcs after maneuvering. The precise orbit determination of the observation arcs before maneuvering and the observation arcs after maneuvering can ensure the convergence of orbit determination because the interference of pulse maneuvers is eliminated. As a result, the orbit determination successfully realizes the maneuver detection, and solves the technical problem in the prior art that when the precise orbit determination of the obtained observation data arc is directly carried out indiscriminately in the case of pulse maneuvers, it is difficult to converge and the orbit determination fails due to divergence. The construction of the maneuvering characteristic parameters in this application is realized by calculating the amount of change in the semi-major axis, eccentricity, and orbital inclination of the orbital element, the previous orbital element, and the subsequent orbital element. This construction implementation makes the observation The errors caused by the short-arc observation of the arc section and the front and rear observation arc sections cancel each other out to a large extent, so that the maneuvering detection under the short-arc observation condition can be realized. Through model design, this application extracts the three key orbit element features of the semi-major axis, eccentricity, and orbital inclination before and after maneuvering of the orbit of the space object, and creatively constructs the characteristic parameters of the maneuver based on this, which is used to characterize the state of motion of the space object. The classic orbit is six miles away, the semi-major axis and eccentricity are sensitive to the component of the pulse maneuver within the orbital plane, and the orbital inclination is sensitive to the component of the impulse maneuver perpendicular to the orbital plane, so the constructed maneuvering characteristic parameters can retain space targets to the greatest extent The maneuvering characteristics of pulse maneuvers are used for learning by the neural network, so as to realize the effective training of the neural network, so that the generated neural network can extract the difference between the characteristic parameters of maneuvering and maneuvering without maneuvering. The target impulsive orbit maneuver detection problem has good generalization performance, thereby avoiding the complex maneuver detection threshold design problem, and improving the calculation efficiency of impulsive orbit maneuver detection compared with traditional methods; The alarm rate further improves the accuracy and accuracy of pulse track maneuver detection.

It should be understood that although the various steps in the flow chart of FIG. 2 are displayed sequentially as indicated by the arrows, these steps are not necessarily executed sequentially in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least some of the steps in FIG. 2 may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but may be executed at different times. The execution of these sub-steps or stages The order is not necessarily performed sequentially, but may be performed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

In a specific embodiment, when there are a large number of space-based optical short-arc observation data arcs of space targets, the artificial intelligence algorithm of the neural network is used to identify the observation arcs before and after the maneuvering of the pulse orbit and carry out the maneuver parameters estimate.

specific:

Assuming that a low-orbit optical observation satellite operating in a sun-synchronous orbit at an orbital height of 800km is used to conduct optical observations of 50 space objects operating in GEO orbits for a period of 7 days, the angular observation error is 3 arc seconds, and the observation start and end time From 2019.12.21.12:00:00 to 2019.12.28.12:00:00, 25 of the satellites had pulsed orbit maneuvers during the observation period, and 712 groups of original observation short-arc segments belonging to different space targets were observed, also known as Observe arcs.

初始时刻的轨道根数为：

。The low-orbit optical observation satellite

The orbital elements at the initial moment are:

.

Since the observation data of actual space targets is difficult to obtain, the training set T and test set D required for neural network training are generated by means of simulation observation. Randomly generate 1,000 simulated space objects in GEO orbit, assuming that the low-orbit optical observation satellites that also operate in a sun-synchronous orbit at an orbital height of 800km are used to conduct a 7-day optical observation of these 1,000 simulated space objects operating in GEO orbit. Observation, the observation start and end time and the orbital element of the low-orbit optical observation satellite are the same as in step 201. Among them, 50% of the simulation targets are selected to randomly add a pulse orbital maneuver within the simulation observation period, and the impulse size of the pulsed orbital maneuver is randomly selected between 2m/s and 5m/s. In order to obtain as many effective learning samples as possible, the application time of the pulse orbital maneuver is randomized within 2019.12.23.00:00:00 to 2019.12.27.00:00:00, and then simulate according to the first four steps (step 201-step 204) Observed and finally obtained about 6547 maneuver characteristic parameters.

Put a maneuver label on each maneuver characteristic parameter, if the moment of applying the pulse maneuver of a certain space target is located in the time interval between the first and second observation arc of the space target or the time interval between the penultimate first and the second observation arc If the number of time intervals or observation arcs is less than 6, since the maneuvering characteristic parameters formed by such spatial objects are difficult to provide effective learning samples for the training of neural networks, such spatial objects should be discarded, leaving 6382 valid samples in the end .

A large number of maneuvering characteristic parameters obtained are randomly divided, and 80% of the sample data is used as the training set T required for neural network training by using the set-out method, and 20% of the sample data is used as the test set D required for neural network training.

Using the generated training set T and test set D, the feedforward neural network model (FNN, Forward NeuralNetwork) is used for supervised training. After testing, when the number of hidden layers is 2 and the number of nodes is 7 and 8, it can achieve relatively good results. good training effect.

The 562 pieces of maneuvering characteristic parameters to be detected after the real observation data are processed through steps 201 to 204 are input into the neural network trained in step 205, and for each input maneuvering characteristic parameter, the trained neural network The corresponding maneuvering labels will be output, and each input maneuvering feature parameter will be corresponded with the corresponding output maneuvering labels one by one and recorded and saved.

The maneuvering characteristic parameters belonging to the same space target are sorted according to the observation time of the first data point of the first observation arc, and then the maneuvering labels of each maneuvering characteristic parameter are detected in turn, and 29 of the 562 maneuvering characteristic parameters are found. The maneuver label marked with 1 by the neural network, according to the maneuver label, the observation arc segment to which the space object belongs is divided into the observation arc segment before the maneuver and the observation arc segment after the maneuver.

，应当认定为无机动目标。Among them, 4 space targets that may have impulsive maneuvers are calculated by track crossing prediction, and the estimated impulse of impulsive maneuvers

, should be identified as a non-maneuvering target.

The test results of the final embodiment are shown in Table 1 below.

Table 1 embodiment test result display table

The present application also provides an orbital mobile intelligent detection device based on short-arc space-based optical observation, as shown in Figure 3, in one embodiment, including: an acquisition module 301, an arrangement module 302, an iteration module 303, and a calculation module 304 , modeling module 305, labeling module 306 and estimation module 307, wherein:

The obtaining module 301 is used to obtain multiple groups of historical observation arcs belonging to different space objects, and uses a quadratic polynomial to fit and screen the historical observation arcs to obtain a preferred arc;

The arrangement module 302 is used to arrange the observation arcs belonging to the same space object in chronological order according to the preferred arcs, and determine the initial orbits of the observation arcs adjacent in time;

The iteration module 303 is used to perform least squares iteration with the initial orbit as an initial value, and perform orbit improvement on observation arcs that belong to the same space target and are adjacent in time, and obtain an orbit improvement result;

The calculation module 304 is used to convert the orbital improvement result into an orbital element, and calculate the changes in the semi-major axis, eccentricity, and orbital inclination of the orbital element, the previous orbital element, and the subsequent orbital element respectively , to obtain the maneuvering characteristic parameters;

Modeling module 305, is used for marking maneuvering label for each characteristic parameter of maneuvering, and according to described maneuvering characteristic parameter and described maneuvering label, neural network is trained, establishes label model;

The label module 306 is used to obtain the current observation arc through short-arc space-based optical observation; according to the current observation arc and the label model, obtain the current maneuvering label of the current observation arc;

The estimation module 307 is used to divide the current observation arcs belonging to the same space object into observation arcs before maneuvering and observation arcs after maneuvering according to the current maneuver label; parameters and complete the intelligent detection of track maneuvering.

For specific limitations on the orbital maneuvering intelligent detection device based on short-arc space-based optical observations, please refer to the above-mentioned limitations on the orbital maneuvering intelligent detection method based on short-arc space-based optical observations, and will not be repeated here. Each module in the above-mentioned device can be fully or partially realized by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, and can also be stored in the memory of the computer device in the form of software, so that the processor can invoke and execute the corresponding operations of the above-mentioned modules.

The technical features of the above embodiments can be combined arbitrarily. To make the description concise, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be It is considered to be within the range described in this specification.

The above-mentioned embodiments only express several implementation modes of the present application, and the description thereof is relatively specific and detailed, but should not be construed as limiting the patent scope of the present application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the scope of protection of the patent application should be based on the appended claims.

Claims (7)

1. The intelligent detection method for the rail maneuvering based on the short arc space-based optical observation is characterized by comprising the following steps:

acquiring a plurality of groups of historical observation arc sections belonging to different space targets, and fitting and screening the historical observation arc sections by using a quadratic polynomial to obtain an optimal arc section; the method specifically comprises the following steps: acquiring a plurality of groups of historical observation arc sections belonging to different space targets, wherein the historical observation arc sections comprise: right ascension and declination; fitting the right ascension and the declination of the historical observation arc section by using a quadratic polynomial to obtain a fitting coefficient of the right ascension and a fitting coefficient of the declination; obtaining the standard deviation of a historical observation arc section according to the fitting coefficient of the right ascension and the fitting coefficient of the declination; screening the historical observation arc sections according to the standard deviation to obtain a preferred arc section;

arranging observation arc sections belonging to the same space target according to the time sequence according to the preferred arc sections, and determining initial tracks of the observation arc sections adjacent in time;

performing least square iteration by taking the initial orbit as an initial value, and performing orbit improvement on the observation arc sections which belong to the same space target and are adjacent in time to obtain an orbit improvement result;

converting the track improvement result into a track number, and respectively calculating the track number, the previous track number and the variation of the next track number in the semimajor axis, the eccentricity and the track inclination angle to obtain a maneuvering characteristic parameter;

marking a maneuvering label for each maneuvering characteristic parameter, training a neural network according to the maneuvering characteristic parameters and the maneuvering labels, and establishing a label model;

obtaining a current observation arc section through short arc space-based optical observation; obtaining a current maneuvering label of the current observation arc section according to the current observation arc section and the label model;

dividing the current observation arc sections belonging to the same space target into a maneuvering front observation arc section and a maneuvering rear observation arc section according to the current maneuvering label; estimating maneuvering parameters and completing intelligent detection of the rail maneuvering according to the maneuvering front observation arc section and the maneuvering rear observation arc section;

according to observation arc section before the maneuver and observation arc section after the maneuver, estimating maneuver parameters and completing the intelligent detection of the track maneuver, specifically: performing least square orbit iteration improvement according to the observation arc section before maneuvering and the observation arc section after maneuvering to obtain an improved result of the orbit before maneuvering and an improved result of the orbit after maneuvering; traversing the track improvement result before maneuvering and the track improvement result after maneuvering within the time interval of the estimated pulse maneuvering application through track cross prediction to obtain the maximum likelihood moment of the pulse maneuvering application; estimating the size and the application direction of the pulse maneuver according to the maximum likelihood moment to finish intelligent detection of the track maneuver;

adopting a quadratic polynomial to fit the right ascension and the declination of the historical observation arc section to obtain a fitting coefficient of the right ascension and a fitting coefficient of the declination, and obtaining a standard deviation of the historical observation arc section according to the fitting coefficient of the right ascension and the fitting coefficient of the declination, wherein the fitting coefficient of the right ascension and the fitting coefficient of the declination are specifically as follows:

and respectively fitting the time-related function expressions of the right ascension and the declination in each historical observation arc segment by adopting a quadratic polynomial, and setting the time-related function expressions of the right ascension alpha and the declination delta as follows:

wherein, a ₀ ，a ₁ ，a ₂ As fitting coefficient of the right ascension, b ₀ ，b ₁ ，b ₂ For the declination fitting coefficient, the initial value of each fitting coefficient is taken as:

due to alpha (t) to a ₀ ，a ₁ ，a ₂ The partial derivatives of (a) are:

therefore, the least square method can be used to obtain the pair a ₀ ，a ₁ ，a ₂ Initial improvement amount delta a ₀ ，Δa ₁ ，Δa ₂ Comprises the following steps:

wherein,

is n _i,j X 3 matrix, B ^T Is a transposed matrix of B, the upper label of the transposed matrix is-1 to represent the inversion operation of the matrix,

is n _i,j The vector of the dimensions is then calculated,

polynomial prediction for right ascension;

then a will be ₀ ，a ₁ ，a ₂ The updating is as follows:

repeating the process of formula (4) and formula (5) up to

Is less than the set threshold value to obtain the fitting coefficient a of the right ascension ₀ ，a ₁ ，a ₂ ；

The same operation steps are carried out on the declination delta to obtain the fitting coefficient b of the declination ₀ ，b ₁ ，b ₂ ；

For each observation time, calculating a right ascension declination fitting value at the corresponding time, and combining the real observation value of the right ascension declination at the corresponding time with the fitting difference to obtain a residual epsilon of the right ascension declination, and further obtaining a standard deviation sigma:

in the formula, x _i Represents the ith residual error, mu is the mean value of the residual errors, and n is the number of data points.

2. The intelligent detection method for track maneuvering based on short arc space-based optical observation according to claim 1, characterized in that the standard deviation of the historical observation arc segment is obtained according to the fitting coefficient of the right ascension and the fitting coefficient of the declination, specifically:

defining an intermediate time of a historical observation arc as

Wherein (int) ((1 + n) _i,j ) /2) intermediate row number representing corresponding observation arc, whereby for each historical observation arc

There is a corresponding intermediate time data point:

wherein alpha is the right ascension at the middle time, delta is the declination at the middle time,

the rate of change of the right ascension at the intermediate time,

the rate of change of declination at the intermediate time,

the position vector and the velocity vector of the optical observation satellite corresponding to the intermediate time are respectively calculated as follows:

3. the intelligent detection method for the track maneuver based on the short-arc space-based optical observation according to claim 1 or 2, characterized in that a maneuver label is marked on each maneuver characteristic parameter, and a neural network is trained according to the maneuver characteristic parameters and the maneuver labels to establish a label model, specifically:

randomly dividing the maneuvering characteristic parameters in proportion to obtain a training set and a test set of the neural network;

printing a maneuvering label for each maneuvering characteristic parameter;

and carrying out supervised training on the neural network according to the training set, the test set and the maneuvering label to establish a label model.

4. The intelligent detection method for the track maneuver based on the short-arc space-based optical observation according to claim 3, wherein a maneuver label is marked on each maneuver characteristic parameter, specifically:

according to the actual maneuvering situation of the corresponding space target, a maneuvering label is marked on each maneuvering characteristic parameter, if the rail maneuvering occurs in a time interval between two observation arc sections of the number of corresponding rails during the calculation of the maneuvering characteristic parameters, the maneuvering characteristic parameters are maneuvering characteristic parameters, and the maneuvering label is set to be 1; otherwise, the maneuvering characteristic parameter is the maneuvering characteristic parameter without maneuvering, and the maneuvering label is set to be 0.

5. The intelligent detection method for the track motor based on the short arc space-based optical observation is characterized in that a current observation arc section is obtained through the short arc space-based optical observation; obtaining a current maneuvering label of the current observation arc section according to the current observation arc section and the label model, wherein the specific steps are as follows:

obtaining a current observation arc section through short arc space-based optical observation;

fitting and screening the current observation arc section by using a quadratic polynomial, determining an initial orbit of the current observation arc section adjacent in time, improving the orbit of the current observation arc section which belongs to the same space target and is adjacent in time, and obtaining a current maneuvering characteristic parameter;

and inputting the current maneuvering characteristic parameters into the label model to obtain the current maneuvering label of the current observation arc section.

6. The intelligent detection method for track maneuvering based on short arc space-based optical observation according to claim 1 or 2, characterized in that, according to the current maneuvering label, the current observation arc segment belonging to the same space target is divided into a maneuvering front observation arc segment and a maneuvering rear observation arc segment, specifically:

sequencing current maneuvering characteristic parameters belonging to the same space target according to the observation time of a first data point of a first observation arc segment of the current maneuvering characteristic parameters, and detecting a current maneuvering label of the current maneuvering characteristic parameters;

if all the current maneuvering tags are 0, the space target does not carry out pulse orbit maneuvering in the whole observation time interval;

if the current maneuvering label is 1, dividing the observation arc section to which the space target belongs into a maneuvering front observation arc section and a maneuvering rear observation arc section by taking two observation arc sections of the number of tracks when the current maneuvering characteristic parameter corresponding to the current maneuvering label is calculated as a boundary point, and defining a time interval between the maneuvering front observation arc section and the maneuvering rear observation arc section as a time interval applied by the estimated pulse maneuvering.

7. Track maneuver intelligent detection device based on short arc sky base optical observation, its characterized in that includes:

the acquisition module is used for acquiring a plurality of groups of historical observation arc sections belonging to different space targets, and fitting and screening the historical observation arc sections by using a quadratic polynomial to obtain an optimal arc section; the method specifically comprises the following steps: acquiring a plurality of groups of historical observation arc sections belonging to different space targets, wherein the historical observation arc sections comprise: right ascension and declination; fitting the right ascension and the declination of the historical observation arc section by using a quadratic polynomial to obtain a fitting coefficient of the right ascension and a fitting coefficient of the declination; obtaining the standard deviation of a historical observation arc section according to the fitting coefficient of the right ascension and the fitting coefficient of the declination; screening the historical observation arc sections according to the standard deviation to obtain an optimal arc section;

the arrangement module is used for arranging the observation arc sections belonging to the same space target according to the preferred arc sections according to a time sequence and determining the initial tracks of the observation arc sections adjacent in time;

the iteration module is used for performing least square iteration by taking the initial orbit as an initial value, and performing orbit improvement on the observation arc sections which belong to the same space target and are adjacent in time to obtain an orbit improvement result;

the calculation module is used for converting the track improvement result into the number of tracks, and respectively calculating the number of tracks, the number of previous tracks and the variation of the number of next tracks in the semimajor axis, the eccentricity and the track inclination angle to obtain maneuvering characteristic parameters;

the modeling module is used for marking each maneuvering characteristic parameter with a maneuvering label, training the neural network according to the maneuvering characteristic parameter and the maneuvering label, and establishing a label model;

the label module is used for obtaining a current observation arc section through short arc space-based optical observation; obtaining a current maneuvering label of the current observation arc section according to the current observation arc section and the label model;

the estimation module is used for dividing the current observation arc sections belonging to the same space target into a before-maneuvering observation arc section and an after-maneuvering observation arc section according to the current maneuvering label; estimating maneuvering parameters and completing intelligent detection of the rail maneuvering according to the maneuvering front observation arc section and the maneuvering rear observation arc section;

according to observation arc section before the maneuver and observation arc section after the maneuver, estimate the maneuver parameter and accomplish the track maneuver intellectual detection, specifically: performing least square orbit iteration improvement according to the observation arc section before maneuvering and the observation arc section after maneuvering to obtain an improvement result of the orbit before maneuvering and an improvement result of the orbit after maneuvering; traversing the track improvement result before maneuvering and the track improvement result after maneuvering within the time interval of the estimated pulse maneuvering application through track cross prediction to obtain the maximum likelihood moment of the pulse maneuvering application; estimating the magnitude and the application direction of the pulse maneuver according to the maximum likelihood moment to finish the intelligent detection of the track maneuver;

adopt quadratic polynomial to the right ascension and the declination of historical observation segmental arc are fitted, obtain the fitting coefficient of right ascension and the fitting coefficient of declination, according to the fitting coefficient of right ascension and the fitting coefficient of declination, obtain the standard deviation of historical observation segmental arc, specifically do:

and respectively fitting the time-related function expressions of the right ascension and the declination in each historical observation arc segment by adopting a quadratic polynomial, and setting the time-related function expressions of the right ascension alpha and the declination delta as follows:

wherein, a ₀ ，a ₁ ，a ₂ Fitting coefficient for the right ascension, b ₀ ，b ₁ ，b ₂ For the declination fitting coefficient, the initial value of each fitting coefficient is taken as:

due to alpha (t) to a ₀ ，a ₁ ，a ₂ The partial derivatives of (a) are:

therefore, the least square method can be used to obtain the pair a ₀ ，a ₁ ，a ₂ Initial improvement Δ a ₀ ，Δa ₁ ，Δa ₂ Comprises the following steps:

wherein,

is n _i,j X 3 matrix, B ^T Is a transposed matrix of B, the upper label-1 represents the inversion operation of the matrix,

is n _i,j The vector of the dimensions is then calculated,

is a polynomial forecast of the right ascension;

then a will be ₀ ，a ₁ ，a ₂ The updating is as follows:

repeating the process of formula (4) and formula (5) until

Is less than the set threshold value to obtain the fitting coefficient a of the right ascension ₀ ，a ₁ ，a ₂ ；

The same operation steps are carried out on the declination delta to obtain declinationFitting coefficient b ₀ ，b ₁ ，b ₂ ；

For each observation time, calculating a right ascension declination fitting value of the corresponding time, and combining the real observation value of the right ascension declination of the corresponding time with the fitting difference to obtain a residual epsilon of the right ascension declination, thereby obtaining a standard deviation sigma:

in the formula, x _i Represents the ith residual error, mu is the mean value of the residual errors, and n is the number of data points.

Images