CN109255393A - Infrared Imaging Seeker anti-jamming performance evaluation method based on random forest - Google Patents

Infrared Imaging Seeker anti-jamming performance evaluation method based on random forest Download PDF

Info

Publication number
CN109255393A
CN109255393A CN201811162639.4A CN201811162639A CN109255393A CN 109255393 A CN109255393 A CN 109255393A CN 201811162639 A CN201811162639 A CN 201811162639A CN 109255393 A CN109255393 A CN 109255393A
Authority
CN
China
Prior art keywords
random forest
infrared imaging
performance evaluation
sample
jamming
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.)
Pending
Application number
CN201811162639.4A
Other languages
Chinese (zh)
Inventor
陆志沣
洪泽华
马潮
杨杰
乔宇
葛辰杰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Institute of Electromechanical Engineering
Original Assignee
Shanghai Institute of Electromechanical Engineering
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shanghai Institute of Electromechanical Engineering filed Critical Shanghai Institute of Electromechanical Engineering
Priority to CN201811162639.4A priority Critical patent/CN109255393A/en
Publication of CN109255393A publication Critical patent/CN109255393A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G06F18/2148—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the process organisation or structure, e.g. boosting cascade
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F41—WEAPONS
    • F41G—WEAPON SIGHTS; AIMING
    • F41G3/00—Aiming or laying means
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/259—Fusion by voting

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The Infrared Imaging Seeker anti-jamming performance evaluation method based on random forest that the present invention provides a kind of, belongs to signal processing and navigational guidance cross-application technical field.Facing to increasingly complicated war environment, the interference free performance of Infrared Imaging Seeker then needs to be continuously improved to cope with this challenge.The appraisal procedure of infra-red missile weapon interference free performance is studied, can be studied for missile weapon system and technical support is provided, be of great significance.The anti-jamming performance evaluation method based on random forest that the invention proposes a kind of, obtains comprehensive interference free performance value under anti-jamming evaluation index system by this method, new thinking is provided for Infrared Imaging Seeker anti-jamming performance evaluation.

Description

Infrared Imaging Seeker anti-jamming performance evaluation method based on random forest
Technical field
The present invention relates between the anti-jamming evaluation index in infrared seeker and anti-disturbance composite performance number fitting and Prediction, for the related data in anti-interference test between infrared seeker and target, is used in a kind of missile guidance field The method that the thought of mathematical analysis carries out quantitative analysis.The invention belongs to signal processings and navigational guidance cross-application technology to lead Domain.
Background technique
Infra-red missile plays an important role on modern battlefield, it has guidance precision height, strong antijamming capability, hidden The advantages that covering property is good, efficiency-cost ratio is high, compact-sized, maneuverability, it has also become one of the precision guided weapon of modern war first choice, In multiple local war, the especially Gulf War and Kosovo War, huge effect [1] has been played.Infrared guidance weapon A large amount of uses, the appearance and fast development of imaging guidance are resulted in, in order to eliminate or reduce infrared guidance guided missile to oneself Various Human disturbance methods are developed actively all to weaken the operation of infrared guidance weapon effect in the threat of square aerial target, countries in the world Energy [1] [2].By development in decades, imaging guidance has also obtained significant progress, the efficiency of infrared guidance guided missile It has been weakened to a certain extent very much.Therefore, effect will very in following war for the weak infra-red missile of anti-interference ability Limited, this makes the interference free performance testing and evaluation of guided missile receive close attention [3].For infrared guidance guided missile, it Operational environment sharply deteriorate, in order to play efficiency in such operational environment, it is desirable that in infrared guidance weapon The performance indicator that target seeker anti-Human disturbance is clearly proposed when development, when the interference free performance index of the guided missile of development meets centainly Condition enables the following guided missile produced under conditions of target aircraft discharges various interference, still is able to one Biggish probability hits the mark, which just has the qualification of batch production.It is therefore desirable in infrared guidance guided missile batch Before production, using the performance indexes in development process, its whole anti-interference can be carried out using suitable method Assessment.
The appraisal procedure of infra-red missile weapon interference free performance and the evaluation index system of foundation can be missile weapon system The very important decision in life cycle management each stage provides technical support, develops and tactics method to missile weapon system planning is improved Scientific, matching equipment construction, in-depth missile armament theories for military operations research and every basic research of development etc. work comprehensively All it is of great significance.
For the anti-jamming performance evaluation of infrared guidance guidance system, there is conflicts at present, are on the one hand due to outer Field target examination will consume a large amount of human and material resources, and the valuableness of every piece of guided missile price leads to that live shell survey can not be carried out in large quantities Examination thus cannot get sufficient sample and carry out statistical estimation；On the other hand be in infra-red missile development process each stage There is a large amount of experimental data that cannot make full use of.Therefore, the anti-interference ability of infrared seeker system how is examined, how to establish one A whole set of science, general anti-jamming performance evaluation index system and simple and effective appraisal procedure has become current infrared Important topic in guidance system evaluation work.
The present invention proposes a kind of Infrared Imaging Seeker anti-jamming performance evaluation method based on random forest, can quantify The every anti-interference index of assessment and target seeker anti-disturbance composite performance number between quantitative relationship, it is anti-for Infrared Imaging Seeker Jamming performance assessment provides new thinking.
Summary of the invention
The present invention is the quantitative relationship in order to establish between anti-jamming evaluation index and anti-disturbance composite performance number, is situated between first Continue the anti-jamming performance evaluation algorithm based on random forest.RF (Random Forest) algorithm is that bagging+ is grown completely The combination of CART tree (post-class processing).It is to establish multiple classification or recurrence mould by bagging method (bootstrapping convergence method) Type is finally used as predicted value using ballot or averagely, can reduce over-fitting.
M wheel is carried out using boostrap (bootstrap) method of sampling to training sample, establishes decision tree respectively, x indicate to Detect sample.Since every wheel is not substantially identical using the sample set gone out, trained model dependency can reduce small.In order into one Step reduces the correlation between model, can carry out stochastical sampling to the feature of training data before every wheel training, can also be in decision Random character selection is carried out on each branch (branch) of tree.
Decision-tree model is a kind of tree structure, the process classified to example or returned based on feature.I.e. according to certain A feature is assigned to Data division several sub-regions (subtree), then divides to subregion recurrence, until meeting some condition then Stopping divides and as leaf node, and the condition that is unsatisfactory for then continues recurrence division.Decision-tree model learning process usually wraps 2 steps It is rapid: the generation of feature selecting, decision tree.
The difference of selection characteristic sequence will generate different decision trees, and selected feature can make label under each subset It is purer.If measures characteristic has drying method, such as error rate, information gain, information gain ratio and Geordie to the quality for generating subset Index etc..
(1) error rate
Training data D by feature A point after several child nodes, select to occur in child node the most class label of number as The return value of this node, is denoted as yc.Then error rate is defined as
Wherein, DcIndicate that the corresponding training data of return value, I indicate to refer to coefficient, yiIndicate the category of i-th of child node Label.
(2) information gain
" comentropy " is a kind of measurement most common index of sample set purity, it is assumed that kth class sample in current sample set D Ratio shared by this is pk, then the comentropy Ent (D) of D is defined as:
Wherein n indicates sample type number.
It is assumed that Category Attributes a has V possible values that can generate V branch if dividing using a to sample D Node, wherein it is a that v-th of branch node, which contains all values on attribute a in D,vSample, be denoted as Dv.According to above formula, meter Calculate DvComentropy, consider further that the sample number that different branch nodes included is different, branch node given to assign weight | Dv |/| D |, i.e. the influence of the more branch node of sample number is bigger, then calculates available attributes a and sample set D is divided " information gain " obtained
In general, information gain is bigger, then mean to carry out dividing using attribute a it is obtained promoted it is bigger.Cause This, can carry out the division Attributions selection of decision tree with information gain.
(3) ratio of profit increase
Actually reference in, information gain criterion for can the more attribute of value number have it is preferred, in order to reduce this The kind possible adverse effect of preference, the not direct use information gain of C4.5 decision tree, but selected most using ratio of profit increase Excellent division attribute, is defined as follows:
Wherein,
It is worth noting that, ratio of profit increase criterion may for can the lesser attribute of value number have preferred, therefore C4.5 Algorithm does not directly select the maximum candidate division attribute of ratio of profit increase instead of, uses a didactic algorithm, first draws from candidate The attribute that information gain is higher than average level is found out in adhering to separately property, and it is highest then therefrom to select ratio of profit increase.
(4) gini index
CART decision tree selects to divide attribute using gini index, and the purity of data set D can be defined as follows with Geordie value:
For intuitive, Gini (D) has been reacted randomly selects two samples from data set, inconsistent general of category label Rate.
Therefore, the gini index of attribute a is defined as follows:
Decision trees are as follows:
Since root node, all possible feature A is calculated on data set D and calculates separately information gain, selects information The maximum feature of gain establishes subset as child node, most to subset as class condition, the value different to this feature respectively Above method is recursively called, until there is no feature can choose or information gain very little.
Next building random forest.In random forest, to each node of base decision tree, first from the category of the node Property set in random selection one include N number of attribute subset, then from this subset select an optimum attributes be used for It divides.Here parameter N controls the introducing degree of randomness.
Compared with prior art, the present invention have it is following the utility model has the advantages that
Random forest is simple, easy to accomplish, computing cost is small, but it shows powerful property in many realistic tasks Energy.It as can be seen that random forest is the improvement to Bagging, but is to pass through with " diversity " of base learner in Bagging Come different and sample disturbance (sampling by initial training collection), the diversity of base learner is not only from sample in random forest Disturbance is also from attribute disturbance, this allows for the Generalization Capability finally integrated can be by the diversity factor between individual learner Increase and further gets a promotion.
Specific embodiment
The present invention is described in detail combined with specific embodiments below.Following embodiment will be helpful to the technology of this field Personnel further understand the present invention, but the invention is not limited in any way.It should be pointed out that the ordinary skill of this field For personnel, without departing from the inventive concept of the premise, several changes and improvements can also be made.These belong to the present invention Protection scope.
Following embodiment is provided in conjunction with the content of the method for the present invention:
Simulation process, which is divided into two steps of model training and model measurement, to carry out.Each sample includes that each assessment refers to when training It is denoted as inputting parameter for multidimensional, whole interference free performance value is as output parameter.
After obtaining final recurrence device, the i.e. exportable whole interference free performance value of anti-interference evaluation index value is inputted.Algorithm One briefly describe it is as shown in table 1:
1 random forests algorithm process of table
Anti-jamming evaluation index, the proper property index and the performance after introducing interference protection measure that can be divided into target seeker change Kind index.As shown in table 3, the proper property index of infrared seeker, including gyro drift rate X1(°/s), minimum distinguishable temperature Poor X2(DEG C), instantaneous field of view X3(×10-7sr)；Performance improvement index after introducing interference protection measure includes discovery real goal Time X4(s), efficiency X is tracked5, tracking accuracy X6(arcsec), operating distance X7(km), anti-Deceiving interference Effective Probability X8With Target image degree of loss X9.Interference free performance value is indicated with Y.
Sample 1-25 is trained using the algorithm based on random forest herein, sample 26-30 is tested, as a result As shown in table 3.The result shows that this method can be obtained about the mapping relations between anti-interference evaluation index and performance number, error Smaller, fitting effect is preferable, has relatively good generalization ability.
2 training sample of table and test sample
3 prediction result of table and error
Serial number Actual value Predicted value Error/% Serial number Actual value Predicted value Error/%
1 0.3799 0.3883 2.2317 16 0.6462 0.5133 20.5621
2 0.5155 0.4992 3.1520 17 0.5404 0.5028 6.9512
3 0.4563 0.4611 1.0664 18 0.3587 0.3933 9.6673
4 0.5306 0.4874 8.1246 19 0.356 0.3726 4.6760
5 0.5205 0.5070 2.5875 20 0.2054 0.3635 77.0011
6 0.485 0.4540 6.3781 21 0.6071 0.5247 13.5592
7 0.4055 0.4162 2.6621 22 0.4505 0.4308 4.3681
8 0.4003 0.4440 10.9386 23 0.3687 0.3945 7.0038
9 0.3295 0.3678 11.6289 24 0.4286 0.4714 10.0042
10 0.5133 0.5108 0.4697 25 0.4678 0.4717 0.84462
11 0.3866 0.4166 7.7758 26 0.2762 0.4307 55.9728
12 0.5519 0.5041 8.6461 27 0.3323 0.3697 11.2822
13 0.3442 0.3900 13.3318 28 0.3886 0.4441 14.2887
14 0.4228 0.4119 2.5725 29 0.4891 0.4134 15.4622
15 0.5445 0.4979 8.5472 30 0.4992 0.4711 5.6227
One skilled in the art will appreciate that in addition to realizing system provided by the invention in a manner of pure computer readable program code It, completely can be by the way that method and step be carried out programming in logic come so that provided by the invention other than system, device and its modules System, device and its modules are declined with logic gate, switch, specific integrated circuit, programmable logic controller (PLC) and insertion The form of controller etc. realizes identical program.So system provided by the invention, device and its modules may be considered that It is a kind of hardware component, and the knot that the module for realizing various programs for including in it can also be considered as in hardware component Structure；It can also will be considered as realizing the module of various functions either the software program of implementation method can be Hardware Subdivision again Structure in part.
Specific embodiments of the present invention are described above.It is to be appreciated that the invention is not limited to above-mentioned Particular implementation, those skilled in the art can make a variety of changes or modify within the scope of the claims, this not shadow Ring substantive content of the invention.In the absence of conflict, the feature in embodiments herein and embodiment can any phase Mutually combination.

Claims (4)

1. a kind of Infrared Imaging Seeker anti-jamming performance evaluation method based on random forest, which is characterized in that input is anti-dry Performance index value output integrated Performance Evaluation value is disturbed, following basic step is specifically included:
Step A: boostrap sampling is carried out to raw data set, generates training set；
Step B: the regression tree of not beta pruning is generated using training set.
Step C: to sample to be tested, anti-disturbance composite performance is exported using random forest.
2. the Infrared Imaging Seeker anti-jamming performance evaluation method according to claim 1 based on random forest, special Sign is that the regression tree that generates is the feature that setting number is randomly selected from training set, and the feature is according to Gini Index chooses optimal characteristics, combines optimal characteristics until regression tree merisis.
3. the Infrared Imaging Seeker anti-jamming performance evaluation method according to claim 2 based on random forest, special Sign is that the Gini index Gini (D) uses:
Wherein, D indicates sample set, pkIndicate ratio shared by kth class sample in sample set D, k=1,2 ... n indicate sample Number.
4. the Infrared Imaging Seeker anti-jamming performance evaluation method according to claim 1 based on random forest, special Sign is that the specific method that the anti-disturbance composite performance calculates is:
Generate M regression tree { hi, i=1 ... M }, M indicates the number of regression tree, hiIndicate i-th of recurrence decision Tree；
Obtain anti-disturbance composite performanceWherein xtIndicate sample to be detected.
CN201811162639.4A 2018-09-30 2018-09-30 Infrared Imaging Seeker anti-jamming performance evaluation method based on random forest Pending CN109255393A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811162639.4A CN109255393A (en) 2018-09-30 2018-09-30 Infrared Imaging Seeker anti-jamming performance evaluation method based on random forest

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811162639.4A CN109255393A (en) 2018-09-30 2018-09-30 Infrared Imaging Seeker anti-jamming performance evaluation method based on random forest

Publications (1)

Publication Number Publication Date
CN109255393A true CN109255393A (en) 2019-01-22

Family

ID=65045361

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811162639.4A Pending CN109255393A (en) 2018-09-30 2018-09-30 Infrared Imaging Seeker anti-jamming performance evaluation method based on random forest

Country Status (1)

Country Link
CN (1) CN109255393A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109726366A (en) * 2018-12-07 2019-05-07 上海机电工程研究所 Infrared Imaging Seeker anti-jamming performance evaluation method, system and medium based on random forest
CN110390400A (en) * 2019-07-02 2019-10-29 北京三快在线科技有限公司 Feature generation method, device, electronic equipment and the storage medium of computation model

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105303262A (en) * 2015-11-12 2016-02-03 河海大学 Short period load prediction method based on kernel principle component analysis and random forest
CN105512768A (en) * 2015-12-14 2016-04-20 上海交通大学 User electricity consumption relevant factor identification and electricity consumption quantity prediction method under environment of big data
CN106372748A (en) * 2016-08-29 2017-02-01 上海交通大学 Hard-rock tunnel boring machine boring efficiency prediction method
CN106600635A (en) * 2016-11-03 2017-04-26 上海机电工程研究所 Infrared target radiation characteristic simulation model checking verifying method based on small subsamples
US20180061091A1 (en) * 2016-08-31 2018-03-01 International Business Machines Corporation Anatomy segmentation through low-resolution multi-atlas label fusion and corrective learning
CN107917646A (en) * 2017-01-10 2018-04-17 北京航空航天大学 A kind of anti-interference method of guidance of strong pulsed D based on the prediction of target terminal accessoble region

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105303262A (en) * 2015-11-12 2016-02-03 河海大学 Short period load prediction method based on kernel principle component analysis and random forest
CN105512768A (en) * 2015-12-14 2016-04-20 上海交通大学 User electricity consumption relevant factor identification and electricity consumption quantity prediction method under environment of big data
CN106372748A (en) * 2016-08-29 2017-02-01 上海交通大学 Hard-rock tunnel boring machine boring efficiency prediction method
US20180061091A1 (en) * 2016-08-31 2018-03-01 International Business Machines Corporation Anatomy segmentation through low-resolution multi-atlas label fusion and corrective learning
CN106600635A (en) * 2016-11-03 2017-04-26 上海机电工程研究所 Infrared target radiation characteristic simulation model checking verifying method based on small subsamples
CN107917646A (en) * 2017-01-10 2018-04-17 北京航空航天大学 A kind of anti-interference method of guidance of strong pulsed D based on the prediction of target terminal accessoble region

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
杨军: "《现代导弹制导控制》", 31 October 2015, 西北工业大学出版社 *
杨照金等: "《工程光学计量测试技术概论》", 29 February 2016, 国防工业出版社 *
郝文宁等: "《数据分析与数据挖掘实验指导书》", 31 March 2016, 国防工业出版社 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109726366A (en) * 2018-12-07 2019-05-07 上海机电工程研究所 Infrared Imaging Seeker anti-jamming performance evaluation method, system and medium based on random forest
CN110390400A (en) * 2019-07-02 2019-10-29 北京三快在线科技有限公司 Feature generation method, device, electronic equipment and the storage medium of computation model
CN110390400B (en) * 2019-07-02 2023-07-14 北京三快在线科技有限公司 Feature generation method and device of computing model, electronic equipment and storage medium

Similar Documents

Publication Publication Date Title
Liu et al. Plant disease recognition: A large-scale benchmark dataset and a visual region and loss reweighting approach
CN109726366A (en) Infrared Imaging Seeker anti-jamming performance evaluation method, system and medium based on random forest
CN109597043A (en) Radar Signal Recognition method based on quantum particle swarm convolutional neural networks
CN109409695B (en) System efficiency evaluation index system construction method and system based on correlation analysis
CN108960330A (en) Remote sensing images semanteme generation method based on fast area convolutional neural networks
CN114564982B (en) Automatic identification method for radar signal modulation type
CN108694390A (en) A kind of cuckoo search improves the modulated signal sorting technique of grey wolf Support Vector Machines Optimized
CN109255393A (en) Infrared Imaging Seeker anti-jamming performance evaluation method based on random forest
CN107330519A (en) Fault Locating Method based on deep neural network
CN109933669B (en) Matching method of battlefield situation data labels
Wang et al. A TPE based inversion of PROSAIL for estimating canopy biophysical and biochemical variables of oilseed rape
McIsaac et al. Using machine learning to autotune chi-squared tests for gravitational wave searches
Dulac et al. Assessing the representation of tropical cyclones in ERA5 with the CNRM tracker
Räty et al. Fusing diameter distributions predicted by an area-based approach and individual-tree detection in coniferous-dominated forests
He et al. MMOS+ ordering search method for Bayesian network structure learning and its application
Luo et al. Semi-supervised deep learning for molecular clump verification
Haron et al. Grading of agarwood oil quality based on its chemical compounds using self organizing map (SOM)
Kahabka ROSAT X-ray sources in the field of the LMC-II. Statistics of background AGN and X-ray binaries
Liu et al. Weight empowerment method in information fusion for radar‐seeker performance evaluation
Komolafe et al. Predictive Modeling for Land Suitability Assessment for Cassava Cultivation
Zheng et al. New incomplete data imputation based on k-nearest neighbor type framework
Santoliquido et al. Classifying binary black holes from Population III stars with the Einstein Telescope: a machine-learning approach
Das Comments on" ANew Algorithm for Generating Prime Implicants"
Tang et al. Radar emitter recognition method based on AdaBoost and decision tree
He et al. A study on evaluation of farmland fertility levels based on optimization of the decision tree 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
RJ01 Rejection of invention patent application after publication

Application publication date: 20190122

RJ01 Rejection of invention patent application after publication