CN106933977A - It is a kind of that the method that flight parameter outlier is rejected in classification is excavated based on big data - Google Patents
It is a kind of that the method that flight parameter outlier is rejected in classification is excavated based on big data Download PDFInfo
- Publication number
- CN106933977A CN106933977A CN201710083694.3A CN201710083694A CN106933977A CN 106933977 A CN106933977 A CN 106933977A CN 201710083694 A CN201710083694 A CN 201710083694A CN 106933977 A CN106933977 A CN 106933977A
- Authority
- CN
- China
- Prior art keywords
- outlier
- data
- flight parameter
- classification
- flight
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2465—Query processing support for facilitating data mining operations in structured databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/17—Function evaluation by approximation methods, e.g. inter- or extrapolation, smoothing, least mean square method
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2216/00—Indexing scheme relating to additional aspects of information retrieval not explicitly covered by G06F16/00 and subgroups
- G06F2216/03—Data mining
Abstract
The method that flight parameter outlier is rejected in classification is excavated based on big data the invention provides a kind of.Aircraft can record flight parameter in flight course in flight recorder, and the outlier for not conforming to convention is included in data, it is necessary to reject the requirement that could meet ex-post analysis.Using data segmentation method by the regular decomposition of flying quality, the characteristics of using curve matching, can Rapid matching relevant data points, and obtain relevant statistical information, the statistical information of different flight parameters is combined into variance group, using radial base neural net classification characteristicses, the flight parameter section comprising outlier can be quickly recognized, then to the position of ill data segment quick identification outlier according to the characteristics of maximin and average, the like, obtain the outlier of the different full data segments of flight parameter.Data of the magnanimity containing wild point, different from the method that outlier is excavated in the filtering of common multiple spot, are especially had preferable rapidity and accuracy by the thought that the method is excavated using big data.
Description
Technical field
The method that flight parameter outlier is rejected in classification is excavated based on big data the invention provides a kind of, belongs to flight control
Data processing field in system, is mainly used in the treatment of unmanned plane during flying data outlier.
Background technology
, it is necessary to various flight parameters in measuring unmanned plane during flying in flight test of unmanned aerial vehicle, and save it in number
According in recorder supply ex-post analysis.But because equipment or the problem of signal, the data of record contain outlier, this outlier in measurement
If not rejecting, considerable influence can be brought for aeroplane performance ex-post analysis, therefore be necessary to take one before data analysis
Fixed method rejects outlier, it is ensured that the completeness and reliability of data.
Outlier method is picked existing, it is general that priori average is calculated using multiple spot smothing filtering, then using currency
With priori average ratio compared with method come judge whether currency be outlier;Or using the method for kalman filtering, estimate priori
The statistical information of value, calculates kalman equations to predict, judges whether currency is outlier.But the above method has amount of calculation
Greatly, the shortcoming that process time is long or computational methods are complex, when flying quality is larger, can spend the more time.
The outlier problem of flying quality how is processed using the thought of data mining, is this area technology urgently to be resolved hurrily
Problem.
The content of the invention
Technology solve problem of the invention is:Overcome the deficiencies in the prior art, there is provided one kind is excavated based on big data and divided
The method that class rejects flight parameter outlier, realizes to unmanned plane magnanimity flying quality rapid extraction outlier.
Technical solution of the invention is:
A kind of method that classification rejecting flight parameter outlier is excavated based on big data is provided, step is as follows:
(1) input matrix P ∈ R are set3×4, wherein column vector is flight parameter, including longitude, latitude, the fitting side of height
Difference, row vector represents four class thresholdings, and respectively each flight parameter is free of outlier, and only longitude contains outlier, and only latitude is containing open country
Value and only height are containing outlier containing the thresholding in the case of outlier;Using radial base neural net (RBN) training matrix P ∈ R3×4;
(2) longitude, latitude and altitude data are read in, each flight parameter data volume is N, by each flight parameter point
It is M one section, is divided into N/M sections;Each flight parameter is fitted per segment data using polynomial fitting method, is intended
Close statistics mean square deviation Matrix C ov_ ∈ R3×(N/M);
(3) the fitting statistics mean square deviation Matrix C ov_ ∈ R for obtaining step (2)3×(N/M)With matrix P ∈ R3×4Carry out thresholding
Inspection, common property gives birth to N/M sort capacity, and they are placed into Class_seri_ ∈ R in class vector group(N/M)×1, realize data segment
Rough segmentation inspection;
(4) the class vector group Class_seri_ ∈ R to being produced in step (3)(N/M)×1It is identified, when telling
When Class_seri_ contains the classification not for first kind thresholding, illustrate that the data segment contains outlier;To longitude, latitude, height point
The data segment containing outlier maximum, minimum value and average max (i), min (i), mean (i), i ∈ [1, N/M] are not extracted into;Meter
CalculateJudge residing scope, whenWhen, by i-th
Maximum max (i) in data in section replaces with average mean (i);WhenWhen,
Minimum value min (i) in data in i-th section is replaced with into average mean (i);When
When, do not processed;After the completion of to the treatment of all data segments, the normal data after rejecting flight parameter outlier is obtained.
Preferably, also include working as in step (4)When, by maximum max (i)
Position charge to outlier record vector WILD_POINT, whenWhen, by minimum value
Min (i) is charged in outlier record vector WILD_POINT.
Preferably, by the position data in WILD_POINT by the method for bubble sort according to arranging from small to large, and will
Deleted with position, obtained the array WILD_POINT that opsition dependent is arranged from small to large.
Preferably, also include being normalized every segment data in step (2), again using multinomial after normalized
Formula approximating method is fitted to each flight parameter per segment data.
Preferably, whenWhen, by the maximum max in the data in i-th section
I () replaces with Kalman Filter Estimation currency out;WhenWhen, by i-th section
In data in minimum value min (i) replace with Kalman Filter Estimation currency out.
Preferably, the method for N/M sort capacity of generation is:By a certain vectorial first fitting variance with only longitude containing open country
Value thresholding contrast, if greater than equal to the threshold value, then the vector is Equations of The Second Kind;If less than the threshold value, then by second
Fitting variance and the contrast of only latitude thresholding containing outlier, if greater than equal to the threshold value, then the vector is the 3rd class;If small
In the threshold value, then the 3rd fitting variance is contrasted with only height containing outlier thresholding containing outlier, if greater than equal to the door
Limit value, then the vector is the 4th class;If less than the threshold value, then the vector is the first kind.
Compared with the prior art, the invention has the advantages that:
(1) present invention processes the outlier problem of flying quality using the thought that big data is excavated, when data volume is huge
It is particularly suitable.It has abandoned prior art needs point-to-point analysis to spend time more shortcoming, using curve matching statistics and god
Fast Classification and the extraction to outlier in big data are realized through the method for network class, open country is excavated by the feature of outlier
Value meets the thought of big data excavation, meets the requirement to the pretreatment of flying quality outlier.
(2) present invention carries out initial data rough sort using neutral net first, then relatively carries out outlier using thresholding
Extract, both taken into account data-handling efficiency, in turn ensure that the accuracy that outlier is extracted.
(3) present invention carries out segment processing to initial data, and every section of calculating variance is contrasted using variance and neutral net
Mode, further improves the disposal ability to big data.
(4) the invention provides the data acquisition system comprising all outlier location points in initial data, consequent malfunction inspection is met
Survey the demand with normal process data.
Brief description of the drawings
Fig. 1 is flow chart of the present invention;
Fig. 2 is flying quality longitude, latitude and height outlier distribution situation;
Fig. 3 is to choose 500 point datas curve (upper figure of the section without outlier longitude, latitude and height fitting of a polynomial:Longitude;
Middle figure:Latitude;Figure below:Highly);
Fig. 4 is curve (the upper figure for choosing 500 point datas section longitude containing outlier, latitude and height fitting of a polynomial:Longitude;
Middle figure:Latitude;Figure below:Highly);
Fig. 5 is selection 500 point datas section outlier classification chart;
Fig. 6 is 500 point datas of selection section by final outlier distribution map after sequence and merging treatment.
Specific embodiment
Unmanned plane realizes navigator fix using GPS in the present invention, in flight course data logger by gps signal and its
He records sensor signal, until when unmanned plane grounds, the gps data that will be recorded in data logger reads out, supplies
Process afterwards and analyze.But in gps signal, the outlier for not conforming to convention is frequently encountered, the presence of which causes aircraft flight
Performance evaluation is in detrimental effect, interference people to the understanding of aeroplane performance, therefore, to gps data before processing, it is necessary to elder generation
One step is processed outlier, meets subsequent analysis demand.
As shown in figure 1, a kind of method that classification rejecting flight parameter outlier is excavated based on big data that the present invention is provided, step
It is rapid as follows:
(1) present invention firstly the need of training RBN neutral nets as longitude and latitude data point matched curve variance high detection door
Limit.Input matrix P ∈ R are set3×4, wherein column vector represent longitude, latitude, height three fitting variances combination, row vector represent
Each the winged thresholding of ginseng containing outlier, totally four groups.Output vector T=[1,2,3,4] ∈ R are set1×4Correspondence column vector difference flies to join and
The type of existing different outlier.Training RBN turns into the thresholding for flying ginseng variance test.
The reliability of thresholding is checked using different characteristic checking R BN, the variance thresholding in input matrix P gusts needs
By the inspection of a large amount of flight parameters, it needs to be free of outlier situation, only longitude in effectively identification longitude and latitude three parameters high
Containing outlier, only latitude is containing outlier and there was only height type containing outlier four, is that the thresholding of all types of settings should be able to be in maximum probability
In the range of recognize that the outlier belongs to that type, when above-mentioned requirements are met, just think RBN training finish, can be used for next
Step detection.
(2) longitude, latitude and altitude supplemental characteristic are read in from the flying quality document downloaded on data logger
DATA, it is assumed that each supplemental characteristic amount is N, is expressed as LON ∈ RN×1,LAT∈RN×1,HIG∈RN×1, by each flight ginseng
Number takes one section by M data point, and N/M sections can be divided into altogether, and every section is represented by LON (i) ∈ RM×1,LAT(i)∈RM×1,HIG(i)
∈RM×1, (i=1...N/M).M data point of every segment data is fitted according to least-squares algorithm using 2 rank multinomials,
The variance that every section is obtained, it is arranged in rows and columns, obtain fitting statistics mean square deviation vector Cov_=[Cov_LON, Cov_LAT, Cov_
HIG]∈R3×(N/M), Cov_ row are comprising warp, latitude, three parameters high to M data point fitting variance, row representative office at same position
Manage N/M times.
(3) by the mean square deviation Vector Groups Cov_ ∈ R of fitting in (2)3×(N/M)Thresholding is checked with the RBN of training in (1) line by line
Contrast, determines the classification often gone respectively, and N/M sort capacity can be produced altogether, and they are placed into Class_ in class vector group
seri_∈R(N/M)×1, the rough segmentation inspection of data segment is realized, in rough segmentation inspection, classification 1 represents three and flies ginseng in same position data segment
The fitting variance of generation is free of outlier in RBN inspection threshold ranges that is, in data segment, and 2,3,4 three kinds of classification are represented respectively
Longitude, each there is the situation of outlier to three kinds of parameters of latitude and height in data segment, in the method, if these three parameters exist
There is outlier simultaneously in same position data segment, then mark classification according to the minimum principle of sequence number sequence number.
(4) the class vector group Class_seri_ ∈ R that will be produced in (3)(N/M)×1It is identified, when telling Class_
When seri_ contains the unit not for 1 type, illustrate that the data segment contains outlier.Related the i-th segment data section containing outlier is extracted
Maximum, minimum value and average max (i), min (i), mean (i), i ∈ [1, N/M] then judge that the relation between them is
It is no to meet formula (1), whereinTo judge the thresholding that outlier whether there is:
When 1 formula in meeting formula (1), outlier is illustrated above average, by maximum point correspondence position record in outlier
In array WILD_POINT, while the corresponding data of this outlier are replaced using average, rejecting flight parameter outlier is obtained
Data;When 3 formula in meeting formula (1), outlier is illustrated below average, by smallest point correspondence position record in outlier number
In group WILD_POINT, while the corresponding data of this outlier are replaced using average;If being unsatisfactory for above-mentioned two formula, number is illustrated
Spread in the high and low thresholds where average according to all values in section i, in the absence of outlier, this data segment is not processed.The present invention
Outlier is replaced using average, it would however also be possible to employ the mode of Kalman filtering estimates the mode of currency, replace outlier.
The method is directed to longitude, and latitude and height three fly each execution of ginseng once, and the section identification of three secondary datas, generation are performed altogether
Outlier record vector WILD_POINT ∈ Rk×1, wherein including k outlier position altogether.
(5) due to during outlier record vector WILD_POINT is produced, being remembered according to maximin location method
Record, therefore position data in WILD_POINT sets not in accordance with ascending order, and also three fly each execution of ginseng
Once this process, all outlier positions are all recorded in same array, it is also possible to have the outlier note of same position
Record, it is therefore desirable to which the outlier for producing WILD_POINT records vector WILD_POINT caving areas, and will remember with position
Record merges deletes, and ultimately generates actual outlier position measuring point WILD_POINT ∈ R(k-p)×1, it is individual identical that p can be deleted among these
Location point, and according to arranging from small to large.
Wherein bubbling method refers to that the current point in array is placed on the 1st point, and is once compared with data thereafter
Compared with when finding that the i-th point data was than the 1st point hour below, then minimum after so follow-up data is traveled through by two data interchanges
Point is then placed on first point, and then current point is placed on the 2nd point, is equally traveled through one time according to upper method, finds the 2nd smallest point
It is placed on position 2, method persistently proceeds to array end to current point according to this, then this array can be arranged according to order from small to large.
But also can be placed in together two same position points by bubbling method, so need again to travel through array one time, find
Data point closely, is merged into one, and array is reduced into size accordingly, until in array without same position
Number, EP (end of program).
(6) array of the WILD_POINT for ultimately generating is exactly the data set comprising all outlier location points in initial data
Close.And then final outlier distribution map is obtained, it is follow-up equipment fault detect or data processing, there is provided foundation.
Flight parameter unruly-value rejecting calculating is carried out using flight control method of the invention, primary condition is remembered for flight parameter
Totally 53040 points, wherein outlier totally 21 are recorded, initial outlier distribution is as shown in Figure 2.
The present invention, as the radial base neural net (RBN) of outlier classification thresholding, sets classification such as table firstly the need of training
Shown in P gusts of 1RBN neutral net threshold parameters and classification T arrange parameter corresponding relations, wherein normal thresholding lat/longitude/height
1/1/3 is corresponded to respectively, and its group indication is 1 class;When only there is longitude outlier, their thresholdings correspond to 1.1/1/3, its point respectively
Class is masked as 2 classes;When only there is latitude outlier, their thresholdings correspond to 1/1.1/3 respectively, and its group indication is 3 classes;When only going out
During existing latitude outlier, their thresholdings correspond to 1/1/3.1 respectively, and its group indication is 4 classes.So by training RBN neutral nets,
Variance thresholding after fitting can be accomplished precise classification.
P gusts of 1 RBN neutral net threshold parameters of table and classification T arrange parameter corresponding relations
P gusts | Normally | Longitude outlier | Latitude outlier | Height outlier |
Longitude | 1 | 1.1 | 1 | 1 |
Latitude | 1 | 1 | 1.1 | 1 |
Highly | 3 | 3 | 3 | 3.1 |
T classifies | 1 | 2 | 3 | 4 |
It is 500 one group to set data segment, carries out polynomial data fitting, finds the side corresponding to lat/longitude/height
Difference Cov_=[Cov_LON, Cov_LAT, Cov_HIG] ∈ R3×106.Without outlier situation 500 point datas fitting as shown in figure 3,
500 point datas fitting containing outlier situation is as shown in Figure 4.
Cov_ is classified by RBN, class vector group Class_seri_ ∈ R can be found106×1, its distribution situation is such as
Shown in Fig. 5, classification 1 is distributed in representing related 500 data point without outlier in figure.
500 data points containing outlier in Class_seri are carried out into maximum, minimum and average identification, it is determined that outlier data
It is stored in outlier array WILD_POINT ∈ R25×1In, then it is ranked up and merges the outlier for ultimately generating and arranging in sequence
Array WILD_POINT ∈ R21×1, its position distribution is as shown in Figure 6.
So, processed by the present invention, the big data containing outlier can quickly position outlier position, meet flying quality afterwards
The need for treatment.
The above, optimal specific embodiment only of the invention, but protection scope of the present invention is not limited thereto,
Any one skilled in the art the invention discloses technical scope in, the change or replacement that can be readily occurred in,
Should all be included within the scope of the present invention.
The content not being described in detail in description of the invention belongs to the known technology of professional and technical personnel in the field.
Claims (6)
1. it is a kind of that the method that flight parameter outlier is rejected in classification is excavated based on big data, it is characterised in that step is as follows:
(1) input matrix P ∈ R are set3×4, wherein column vector be flight parameter, including longitude, latitude, height fitting variance,
Row vector represents four class thresholdings, and respectively each flight parameter is free of outlier, and only longitude contains outlier, only latitude containing outlier and
Only height contains the thresholding in the case of outlier containing outlier;Using radial base neural net (RBN) training matrix P ∈ R3×4;
(2) longitude, latitude and altitude data are read in, each flight parameter data volume is N, and each flight parameter is divided into M
Individual one section, it is divided into N/M sections;Each flight parameter is fitted per segment data using polynomial fitting method, is fitted
Statistics mean square deviation Matrix C ov_ ∈ R3×(N/M);
(3) the fitting statistics mean square deviation Matrix C ov_ ∈ R for obtaining step (2)3×(N/M)With matrix P ∈ R3×4Carry out threshold test,
Common property gives birth to N/M sort capacity, and they are placed into class vector group Class_seri_ ∈ R(N/M)×1In, realize the thick of data segment
Go-on-go;
(4) the class vector group Class_seri_ ∈ R to being produced in step (3)(N/M)×1It is identified, when telling Class_
When seri_ contains the classification not for first kind thresholding, illustrate that the data segment contains outlier;Longitude, latitude, height will be contained respectively
The data segment of outlier extracts maximum, minimum value and average max (i), min (i), mean (i), i ∈ [1, N/M];CalculateJudge residing scope, whenWhen, by i-th section
In data in maximum max (i) replace with average mean (i);WhenWhen,
Minimum value min (i) in data in i-th section is replaced with into average mean (i);When
When, do not processed;After the completion of to the treatment of all data segments, the normal data after rejecting flight parameter outlier is obtained,It is threshold
Value.
2. it is as claimed in claim 1 that the method that flight parameter outlier is rejected in classification is excavated based on big data, it is characterised in that step
Suddenly also include working as in (4)When, outlier record is charged into the position of maximum max (i)
In vectorial WILD_POINT, whenWhen, minimum value min (i) is charged into outlier record
In vectorial WILD_POINT.
3. it is as claimed in claim 2 that the method that flight parameter outlier is rejected in classification is excavated based on big data, it is characterised in that will
Position data in WILD_POINT according to arranging from small to large, and will be deleted by the method for bubble sort with position, be obtained
To the array WILD_POINT that opsition dependent is arranged from small to large.
4. it is as claimed in claim 1 that the method that flight parameter outlier is rejected in classification is excavated based on big data, it is characterised in that step
Suddenly also include being normalized every segment data in (2), again using polynomial fitting method to each after normalized
Flight parameter is fitted per segment data.
5. it is as claimed in claim 1 that the method that flight parameter outlier is rejected in classification is excavated based on big data, it is characterised in that step
Suddenly in (4), whenWhen, maximum max (i) in the data in i-th section is replaced with
Kalman Filter Estimation currency out;WhenWhen, by the data in i-th section
In minimum value min (i) replace with Kalman Filter Estimation currency out.
6. it is as claimed in claim 1 that the method that flight parameter outlier is rejected in classification is excavated based on big data, it is characterised in that to produce
The method of raw N/M sort capacity is:By a certain vectorial first fitting variance and the contrast of only longitude thresholding containing outlier, if greatly
In equal to the threshold value, then the vector is Equations of The Second Kind;If less than the threshold value, then by second fitting variance and only latitude
Thresholding containing outlier is contrasted, and if greater than equal to the threshold value, then the vector is the 3rd class;If less than the threshold value, then by
Three fitting variances are contrasted with only height containing outlier thresholding containing outlier, and if greater than equal to the threshold value, then the vector is the
Four classes;If less than the threshold value, then the vector is the first kind.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710083694.3A CN106933977B (en) | 2017-02-16 | 2017-02-16 | A method of classification is excavated based on big data and rejects flight parameter outlier |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710083694.3A CN106933977B (en) | 2017-02-16 | 2017-02-16 | A method of classification is excavated based on big data and rejects flight parameter outlier |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106933977A true CN106933977A (en) | 2017-07-07 |
CN106933977B CN106933977B (en) | 2019-08-09 |
Family
ID=59423246
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710083694.3A Active CN106933977B (en) | 2017-02-16 | 2017-02-16 | A method of classification is excavated based on big data and rejects flight parameter outlier |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106933977B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111736626A (en) * | 2020-06-22 | 2020-10-02 | 中国人民解放军国防科技大学 | Stable missile path data processing method |
CN111815806A (en) * | 2020-07-10 | 2020-10-23 | 中国人民解放军空军工程大学 | Method for preprocessing flight parameter data based on wild value elimination and feature extraction |
CN111881213A (en) * | 2020-07-28 | 2020-11-03 | 东航技术应用研发中心有限公司 | System for storing, processing and using flight big data |
CN112965964A (en) * | 2021-02-08 | 2021-06-15 | 北京军懋国兴科技股份有限公司 | Wild value detection method and system for actually measured flight parameter data and computer related product |
CN113933876A (en) * | 2021-11-16 | 2022-01-14 | 西南电子技术研究所(中国电子科技集团公司第十研究所) | Multi-satellite communication time difference positioning data fusion processing method |
CN114219034A (en) * | 2021-12-17 | 2022-03-22 | 江西洪都航空工业集团有限责任公司 | SVM classification algorithm-based flight parameter data and pilot physiological data mining method |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6882958B2 (en) * | 2001-06-28 | 2005-04-19 | National Instruments Corporation | System and method for curve fitting using randomized techniques |
CN102650527A (en) * | 2012-05-25 | 2012-08-29 | 北京航空航天大学 | Temperature compensation method for denoising fiber-optic gyroscope on basis of time series analysis |
CN104700153A (en) * | 2014-12-05 | 2015-06-10 | 江南大学 | PH (potential of hydrogen) value predicting method of BP (back propagation) neutral network based on simulated annealing optimization |
CN106384080A (en) * | 2016-08-31 | 2017-02-08 | 广州精点计算机科技有限公司 | Apparent age estimating method and device based on convolutional neural network |
-
2017
- 2017-02-16 CN CN201710083694.3A patent/CN106933977B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6882958B2 (en) * | 2001-06-28 | 2005-04-19 | National Instruments Corporation | System and method for curve fitting using randomized techniques |
CN102650527A (en) * | 2012-05-25 | 2012-08-29 | 北京航空航天大学 | Temperature compensation method for denoising fiber-optic gyroscope on basis of time series analysis |
CN104700153A (en) * | 2014-12-05 | 2015-06-10 | 江南大学 | PH (potential of hydrogen) value predicting method of BP (back propagation) neutral network based on simulated annealing optimization |
CN106384080A (en) * | 2016-08-31 | 2017-02-08 | 广州精点计算机科技有限公司 | Apparent age estimating method and device based on convolutional neural network |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111736626A (en) * | 2020-06-22 | 2020-10-02 | 中国人民解放军国防科技大学 | Stable missile path data processing method |
CN111736626B (en) * | 2020-06-22 | 2022-05-17 | 中国人民解放军国防科技大学 | Stable missile path data processing method |
CN111815806A (en) * | 2020-07-10 | 2020-10-23 | 中国人民解放军空军工程大学 | Method for preprocessing flight parameter data based on wild value elimination and feature extraction |
CN111815806B (en) * | 2020-07-10 | 2021-11-16 | 中国人民解放军空军工程大学 | Method for preprocessing flight parameter data based on wild value elimination and feature extraction |
CN111881213A (en) * | 2020-07-28 | 2020-11-03 | 东航技术应用研发中心有限公司 | System for storing, processing and using flight big data |
CN112965964A (en) * | 2021-02-08 | 2021-06-15 | 北京军懋国兴科技股份有限公司 | Wild value detection method and system for actually measured flight parameter data and computer related product |
CN112965964B (en) * | 2021-02-08 | 2023-09-08 | 北京军懋国兴科技股份有限公司 | Wild value detection method and system for actually measured flight parameter data and computer related product |
CN113933876A (en) * | 2021-11-16 | 2022-01-14 | 西南电子技术研究所(中国电子科技集团公司第十研究所) | Multi-satellite communication time difference positioning data fusion processing method |
CN113933876B (en) * | 2021-11-16 | 2023-05-23 | 西南电子技术研究所(中国电子科技集团公司第十研究所) | Multi-star communication time difference positioning data fusion processing method |
CN114219034A (en) * | 2021-12-17 | 2022-03-22 | 江西洪都航空工业集团有限责任公司 | SVM classification algorithm-based flight parameter data and pilot physiological data mining method |
Also Published As
Publication number | Publication date |
---|---|
CN106933977B (en) | 2019-08-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106933977B (en) | A method of classification is excavated based on big data and rejects flight parameter outlier | |
CN104931960B (en) | Trend message and radar target situation information full flight path segment data correlating method | |
CN106680798B (en) | A kind of identification of airborne LIDAR air strips overlay region redundancy and removing method | |
CN103390169A (en) | Sorting method of vehicle-mounted laser scanning point cloud data of urban ground objects | |
CN103366602A (en) | Method of determining parking lot occupancy from digital camera images | |
CN109050535A (en) | A kind of quick landform industry and mining city method based on vehicle attitude | |
CN107688782A (en) | Oil tank detection and reserve analysis method based on high-resolution optical remote sensing image | |
CN106897681A (en) | A kind of remote sensing images comparative analysis method and system | |
CN106408030A (en) | SAR image classification method based on middle lamella semantic attribute and convolution neural network | |
CN103575279B (en) | Based on Data Association and the system of fuzzy message | |
CN108875805A (en) | The method for improving detection accuracy using detection identification integration based on deep learning | |
CN104103070B (en) | Landing point selecting method based on optical images | |
CN107657224A (en) | A kind of multilayer parallel network SAR image Aircraft Targets detection method based on part | |
CN109344812A (en) | A kind of improved single photon point cloud data denoising method based on cluster | |
CN108921219A (en) | Model identification method based on target track | |
CN104269057A (en) | Bayonet sensor layout method based on floating car OD data | |
CN110196962A (en) | A kind of air speed abnormality recognition method based on Density Estimator | |
CN104008403B (en) | A kind of SVM(Vector machine)The multi-targets recognition decision method of pattern | |
CN102663740B (en) | SAR image change detection method based on image cutting | |
CN111199243B (en) | Aerial target identification method and system based on improved decision tree | |
CN112949552A (en) | Fall detection processing method and device | |
Sirat et al. | Comparative analysis of K-means and isodata algorithms for clustering of fire point data in Sumatra region | |
CN107132515A (en) | A kind of point mark screening technique constrained based on multidimensional information | |
CN105118332B (en) | A kind of air traffic control analog simulation method for detecting abnormality and device based on clustering methodology | |
CN106951924B (en) | Seismic coherence body image fault automatic identification method and system based on AdaBoost algorithm |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |