CN108734226A - Decision fusion method, apparatus and system - Google Patents
Decision fusion method, apparatus and system Download PDFInfo
- Publication number
- CN108734226A CN108734226A CN201810600887.6A CN201810600887A CN108734226A CN 108734226 A CN108734226 A CN 108734226A CN 201810600887 A CN201810600887 A CN 201810600887A CN 108734226 A CN108734226 A CN 108734226A
- Authority
- CN
- China
- Prior art keywords
- sensor
- decision
- target category
- probability
- reliability
- 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
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Testing Or Calibration Of Command Recording Devices (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
A kind of Decision fusion method, apparatus of present invention offer and system, belong to Decision fusion technical field, can solve the problems, such as that existing Decision fusion result is inaccurate.The Decision fusion method of the present invention, which is characterized in that including：The target category set of sensor is obtained, target category set includes c kind target categories, c>1；The observation data and centrad of sensor are obtained, and determine the local decision-making classification of sensor according to observation data；According to the observation data, centrad and c target category of sensor, according to the first preset algorithm, the reliability of sensor is determined respectively；The decision probability of sensor is determined according to the observation data of sensor and c target category；Wherein, decision probability is the conditional probability that the local decision-making classification of sensor is respectively different target classification；The global decisions of sensor are determined according to the reliability of sensor and decision probability.
Description
Technical field
The invention belongs to Decision fusion technical fields, and in particular to a kind of Decision fusion method, apparatus and system.
Background technology
With the development of sensor technology, the calculating of sensor and storage capacity are greatly improved.Decision fusion method
Node (sensor) is allowed to be converted to local decision-making using the data that the ability of itself will be observed that and upload to fusion center,
Eventually become a global decisions.In Decision fusion method, since data volume is greatly reduced, transmission quantity is consumed
Energy greatly reduce, to improve the life span of network.
In wireless sensor network, node-node transmission finite energy and transmission bandwidth relatively supports, under many scenes directly
It is undesirable to server transmission data.Therefore it needs to carry out Classification and Identification to data in local using Decision fusion method, and
Local decision-making result is sent to fusion center to merge, to eliminate the limitation of energy and transmission bandwidth, to be promoted
Global recognition precision.Therefore, how under the premise of limiting certain volume of transmitted data, the precision tool of Decision fusion result is promoted
It is significant.
Currently used Decision fusion method mainly has：Probability fusion method and rough set fusion method.And wherein, probability
The ambiguity and accuracy of observation data can not be effectively treated in fusion method, and probability fusion method needs to understand observation number in advance
According to prior distribution, this can not be predicted in advance in most cases, therefore have certain limitation；Rough set merges
The basic thought of method is the division that problem is counted as domain, and the level of detail of data is indicated with " granularity ".When
When granularity is bigger, problem just will appear uncertainty.Redundant attributes can be eliminated by attribute loop process, and are merged
As a result.It is currently also fewer in practical applications but since rough set fusion method compares forward position.
In short, or current Decision fusion method needs know the prior distribution of data or in practical application in advance
In it is also fewer, cause the waste of the inexactness of Decision fusion result and the transmission quantity of data.
Invention content
The present invention is directed at least solve one of the technical problems existing in the prior art, it one kind is provided can improve decision and melt
Close the Decision fusion method of precision.
Technical solution is a kind of Decision fusion method used by solving present invention problem, including：
The target category set of sensor is obtained, the target category set includes c kind target categories, c>1；
The observation data and centrad of sensor are obtained, and determine the local decision-making class of sensor according to the observation data
Not；
It is determined respectively according to the first preset algorithm according to the observation data, centrad and c target category of sensor
The reliability of sensor；
The decision probability of sensor is determined according to the observation data of sensor and c target category；Wherein, the decision is general
Rate is that the local decision-making classification of sensor is respectively the conditional probability of different target classification；
The global decisions of sensor are determined according to the reliability and decision probability of sensor.
Preferably, the step of target category set for obtaining sensor includes：
The conception of history measured data for obtaining sensor, corresponding target category is determined according to the conception of history measured data, is generated
Target category set.
Preferably, first preset algorithm is specially：Wherein, r is sensor
Reliability, λ are the centrad of sensor, and β is constant,It is in local decision-making classification and the target category set of sensor the
The Euclidean distance of j target category.
Preferably, the described the step of global decisions of sensor are determined according to the reliability and decision probability of sensor
It specifically includes：
It is merged according to the second preset algorithm according to the conditional probability of sensor and reliability, obtains sensor
Global decisions；
Wherein, second preset algorithm is specially：Wherein, ω is that the overall situation of sensor is determined
Plan, P are the decision probability of sensor.
Technical solution is a kind of Decision fusion device used by solving present invention problem, including：
Target category acquiring unit, the target category set for obtaining sensor, the target category set include c
Target category, c>1；
Local decision-making determination unit, observation data and centrad for obtaining sensor, and according to the observation data
Determine the local decision-making classification of sensor；
Reliability determination unit, for according to the observation data, centrad and c target category of sensor, according to the
One preset algorithm determines the reliability of sensor respectively；
Decision probability determination unit, for determining determining for sensor according to the observation data and c target category of sensor
Plan probability；Wherein, the decision probability is the conditional probability that the local decision-making classification of sensor is respectively different target classification；
Global decisions unit, for determining that the overall situation of sensor is determined according to the reliability and decision probability of sensor
Plan.
Preferably, the target category acquiring unit is specifically used for, and the conception of history measured data of sensor is obtained, according to described
Conception of history measured data determines corresponding target category.
Preferably, first preset algorithm is specially：Wherein, r is sensor
Reliability, λ are the centrad of sensor, and β is constant,Local decision-making classification for sensor and jth in target category set
The Euclidean distance of a target category.
Preferably, the global decisions unit is specifically used for,
It is merged according to the second preset algorithm according to the conditional probability of sensor and reliability, obtains sensor
Global decisions；
Wherein, second preset algorithm is specially：Wherein, ω is that the overall situation of sensor is determined
Plan, P are the decision probability of sensor.
Technical solution is a kind of Decision fusion system used by solving present invention problem, including：
Any one of the above Decision fusion device；
Sensor；
The Decision fusion device is used to carry out Decision fusion according to the observation data of the sensor.
In the Decision fusion method of the present invention, the target category of sensor, and base are obtained using the historical data of sensor
In the calculating reliability of target category and the other distance of Decision Classes and the centrad of sensor, reliability and conditional probability are carried out
Quantization and fusion, so as to effectively improve the precision of Decision fusion result.
Description of the drawings
Fig. 1 is the flow chart of the Decision fusion method of the embodiment of the present invention 1；
Fig. 2 is the schematic diagram of the Decision fusion device of the embodiment of the present invention 2.
Specific implementation mode
To make those skilled in the art more fully understand technical scheme of the present invention, below in conjunction with the accompanying drawings and specific embodiment party
Present invention is further described in detail for formula.
Embodiment 1：
As shown in Figure 1, the present embodiment provides a kind of Decision fusion method, it can be used for the history observation number according to sensor
Quantitative fusion is carried out according to Current observation data, to improve the precision of the Decision fusion result of sensor.
The Decision fusion method includes：
S11, the target category set for obtaining sensor, target category set includes c kind target categories, c>1.
Wherein, target category refers to after each observation, the decision classification that may recognize that according to the observation data of sensor.It passes
The observation data that repeatedly observation obtains, according to different observation data, the different targets that can obtain are locally stored in sensor
Classification, to form target category set θ={ H1,,,Hc, wherein HcFor c-th of target category.
Specifically, obtain sensor k target category set when, can by obtain sensor k conception of history measured data,
Corresponding target category is determined according to the conception of history measured data of sensor k.Wherein, conception of history measured data is that sensor is determined at this
The observation data that repeatedly observation obtains before plan fusion, the target category of sensor is determined by these conception of history measured data, from
And keep the target category obtained more acurrate.
S12, the observation data and centrad for obtaining sensor, and determine the local decision-making class of sensor according to observation data
Not.
In this step, the local decision-making classification u of sensor is obtained by obtaining the observation data x of sensor.Wherein, by
In observe data be sensor single observation as a result, observation process and determine local decision-making classification during can exist
Certain error, therefore the local decision-making classification of the sensor obtained in this step can have certain miss with practical right decision classification
Difference.
Meanwhile in the present embodiment, the centrad λ of sensor can be also obtained, which indicates in total network links, periphery
Other sensors to the degree of support of the sensor.Wherein, it is independent of one another with centrad due to the observation data of sensor
, therefore the two can also be separated and successively be obtained.
S13, observation data, centrad and c target category according to sensor are determined according to the first preset algorithm
The reliability of sensor.
In this step, in conjunction with observation data x, centrad λ and the target category of sensor, the reliability of sensor is determined
R show that global decisions provide basis for the Decision fusion in subsequent step.
Wherein, it is preferred that the first preset algorithm is specially：Wherein, r is sensor
Reliability, λ are the centrad of sensor, and β is constant,It is in local decision-making classification and the target category set of sensor the
The Euclidean distance of j target category.For sensor k, then the first preset algorithm isIts
In, rkFor the reliability of sensor k, λkFor the centrad of sensor k, β is constant,For the local decision-making class of sensor k
Other ukWith target category set { H1,,,HcIn j-th of target category (target category Hj) Euclidean distance.That is, this step
In, it is counted respectively by the different target classification in the observation data, centrad and target category set of comprehensive sensor k
It calculates, therefrom selects reliability of the maximum result of calculation of numerical value as the sensor.
S14, the decision that sensor is determined according to c target category in the observation data and target category set of sensor
Probability.
Wherein, decision probability be sensor local decision-making classification be respectively different target classification conditional probability collection
It closes, specifically, the decision probability P of sensor kk={ pk(uk|H1), pk(uk|H2),...,pk(uk|Hc)}.Wherein, sensor k
Decision classification ukFor target category HjConditional probabilityWherein, pk(Hj) be target category it is Hj
Probability, pk(ukHj) it is to determine that the local decision-making classification of sensor k is current local decision-making classification uk, and target category is Hj
Probability.pk(Hj) and pk(ukHj) can be obtained by the history observational data statistical of sensor k, such as according to sensor k's
It is respectively H that conception of history measured data, which counts target category,1,H2,H3,...,HcNumber, to show that sensor target classification is
HjProbability.
S15, the global decisions that sensor is determined according to the reliability and decision probability of sensor.
Preferably, this step specifically includes：According to the conditional probability of sensor and reliability according to the second preset algorithm into
Row fusion, obtains the global decisions of sensor.Wherein, the second preset algorithm is specially：Wherein, ω
For the global decisions of sensor, P is the decision probability of sensor.For sensor k, then the second preset algorithm is：Wherein, ωkFor the global decisions of sensor k, PkFor the decision probability of sensor k.That is, this
In step, sensor-based each conditional probability is merged with reliability, and the wherein maximum fusion results of numerical value are as final
Global decisions.In this step, since reliability is the centrad based on target category Yu Decision Classes other distance and sensor
It is calculated, therefore when being quantified and being merged using reliability and conditional probability, Decision fusion result can be effectively improved
Precision.
In Decision fusion method provided in this embodiment, the target class of sensor is obtained using the historical data of sensor
Not, and the calculating reliability of the centrad based on target category and the other distance of Decision Classes and sensor, to reliability and condition
Probability is quantified and is merged, so as to effectively improve the precision of Decision fusion result.
Embodiment 2：
As shown in Fig. 2, the present embodiment provides a kind of Decision fusion device, it can be used for the conception of history measured data according to sensor
Quantitative fusion is carried out to Current observation data, to improve the precision of the Decision fusion result of sensor, with to passing in the present embodiment
Sensor k illustrated for Decision fusion.
The Decision fusion device includes：Target category acquiring unit, local decision-making determination unit, reliability determination unit,
Decision probability determination unit and global decisions unit.Wherein,
Target category acquiring unit is used to obtain the target category of sensor k, forms target category set, target category collection
Conjunction includes c target category, c>1.Preferably, in the present embodiment, target category acquiring unit is specifically used for, and obtains sensor k
Conception of history measured data, corresponding with sensor k target category is determined according to the conception of history measured data of sensor k.
Local decision-making determination unit is used to obtain the observation data and centrad of sensor k, and is determined according to observation data
The local decision-making classification of sensor k.
Reliability determination unit, for according to the observation data, centrad and c target category of sensor k, according to the
One preset algorithm determines the reliability of sensor k respectively.Wherein, the first preset algorithm is concretely：Wherein, rkFor the reliability of sensor k, λkFor the centrad of sensor k, β is constant,
For the local decision-making classification u of sensor kkWith target category set { H1,...,HcIn j-th of target category (target category Hj)
Euclidean distance.
Decision probability determination unit is used to determine determining for sensor k according to the observation data and c target category of sensor k
Plan probability；Wherein, decision probability is the conditional probability that the local decision-making classification of sensor k is respectively different target classification.
Global decisions unit, the global decisions for determining sensor k according to the reliability and decision probability of sensor k.
Preferably, global decisions unit is specifically used for, and is carried out according to the second preset algorithm according to the conditional probability of sensor k and reliability
Fusion, obtains the global decisions of sensor k；Wherein, the second preset algorithm is specially：Wherein,
ωkFor the global decisions of sensor k, PkFor the decision probability of sensor k.
In Decision fusion device provided in this embodiment, the target class of sensor is obtained using the historical data of sensor
Not, and the calculating reliability of the centrad based on target category and the other distance of Decision Classes and sensor, to reliability and condition
Probability is quantified and is merged, so as to effectively improve the precision of Decision fusion result.
Embodiment 3：
The present embodiment provides a kind of Decision fusion systems, including：Decision fusion device described in embodiment 2 and sensing
Device, wherein Decision fusion device is used to carry out Decision fusion according to the observation data of sensor.
In Decision fusion system provided in this embodiment, Decision fusion device obtains sensing using the historical data of sensor
The target category of device, and the calculating reliability of the centrad based on target category and the other distance of Decision Classes and sensor, pair can
Quantified and merged by degree and conditional probability, so as to effectively improve the precision of Decision fusion result.
It is understood that the principle that embodiment of above is intended to be merely illustrative of the present and the exemplary implementation that uses
Mode, however the present invention is not limited thereto.For those skilled in the art, in the essence for not departing from the present invention
In the case of refreshing and essence, various changes and modifications can be made therein, these variations and modifications are also considered as protection scope of the present invention.
Claims (9)
1. a kind of Decision fusion method, which is characterized in that including：
The target category set of sensor is obtained, the target category set includes c kind target categories, c>1；
The observation data and centrad of sensor are obtained, and determine the local decision-making classification of sensor according to the observation data；
According to the observation data, centrad and c target category of sensor, according to the first preset algorithm, sensing is determined respectively
The reliability of device；
The decision probability of sensor is determined according to the observation data of sensor and c target category；Wherein, the decision probability is
The local decision-making classification of sensor is respectively the conditional probability of different target classification；
The global decisions of sensor are determined according to the reliability and decision probability of sensor.
2. Decision fusion method according to claim 1, which is characterized in that the target category set for obtaining sensor
The step of include：
The conception of history measured data for obtaining sensor determines corresponding target category according to the conception of history measured data, generates target
Category set.
3. Decision fusion method according to claim 1, which is characterized in that
First preset algorithm is specially：Wherein, r is the reliability of sensor, and λ is to pass
The centrad of sensor, β are constant,Local decision-making classification for sensor and j-th of target category in target category set
Euclidean distance.
4. Decision fusion method according to claim 1, which is characterized in that
Described the step of determining the global decisions of sensor according to the reliability and decision probability of sensor, specifically includes：
It is merged according to the second preset algorithm according to the conditional probability of sensor and reliability, obtains the overall situation of sensor
Decision；
Wherein, second preset algorithm is specially：Wherein, ω is the global decisions of sensor, and P is
The decision probability of sensor.
5. a kind of Decision fusion device, which is characterized in that including：
Target category acquiring unit, the target category set for obtaining sensor, the target category set include c target
Classification, c>1；
Local decision-making determination unit, observation data and centrad for obtaining sensor, and determined according to the observation data
The local decision-making classification of sensor；
Reliability determination unit, it is pre- according to first for the observation data, centrad and c target category according to sensor
Imputation method determines the reliability of sensor respectively；
Decision probability determination unit, for determining that the decision of sensor is general according to the observation data and c target category of sensor
Rate；Wherein, the decision probability is the conditional probability that the local decision-making classification of sensor is respectively different target classification；
Global decisions unit, the global decisions for determining sensor according to the reliability and decision probability of sensor.
6. Decision fusion device according to claim 5, which is characterized in that the target category acquiring unit is specifically used
In obtaining the conception of history measured data of sensor, corresponding target category determined according to the conception of history measured data.
7. Decision fusion device according to claim 5, which is characterized in that
First preset algorithm is specially：Wherein, r is the reliability of sensor, and λ is to pass
The centrad of sensor, β are constant,For j-th target category in local decision-making classification and the target category set of sensor
Euclidean distance.
8. Decision fusion device according to claim 5, which is characterized in that
The global decisions unit is specifically used for,
It is merged according to the second preset algorithm according to the conditional probability of sensor and reliability, obtains the overall situation of sensor
Decision；
Wherein, second preset algorithm is specially：Wherein, ω is the global decisions of sensor, and P is
The decision probability of sensor.
9. a kind of Decision fusion system, which is characterized in that including：
Decision fusion device described in any one of claim 5 to 8；
Sensor；
The Decision fusion device is used to carry out Decision fusion according to the observation data of the sensor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810600887.6A CN108734226A (en) | 2018-06-12 | 2018-06-12 | Decision fusion method, apparatus and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810600887.6A CN108734226A (en) | 2018-06-12 | 2018-06-12 | Decision fusion method, apparatus and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108734226A true CN108734226A (en) | 2018-11-02 |
Family
ID=63929335
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810600887.6A Pending CN108734226A (en) | 2018-06-12 | 2018-06-12 | Decision fusion method, apparatus and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108734226A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102829780A (en) * | 2012-08-30 | 2012-12-19 | 西安电子科技大学 | X-ray pulsar weak signal detection method based on decision information fusion |
| CN103985381A (en) * | 2014-05-16 | 2014-08-13 | 清华大学 | Voice frequency indexing method based on parameter fusion optimized decision |
| CN105975744A (en) * | 2016-04-22 | 2016-09-28 | 西安工程大学 | D-S evidence theory-based textile process data fusion system |
-
2018
- 2018-06-12 CN CN201810600887.6A patent/CN108734226A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102829780A (en) * | 2012-08-30 | 2012-12-19 | 西安电子科技大学 | X-ray pulsar weak signal detection method based on decision information fusion |
| CN103985381A (en) * | 2014-05-16 | 2014-08-13 | 清华大学 | Voice frequency indexing method based on parameter fusion optimized decision |
| CN105975744A (en) * | 2016-04-22 | 2016-09-28 | 西安工程大学 | D-S evidence theory-based textile process data fusion system |
Non-Patent Citations (3)
| Title |
|---|
| WENYU ZHANG 等: "Belief Function Based Decision Fusion for Decentralized Target Classification in Wireless Sensor Networks", 《SENSORS》 * |
| 严志军: "基于无线传感器网络的多源数据融合技术研究", 《中国优秀硕士学位论文全文数据库》 * |
| 张文宇 等: "基于证据理论的无线传感器网络决策融合算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20140160295A1 (en) | Road condition detection | |
| TWI709932B (en) | Method, device and equipment for monitoring transaction indicators | |
| WO2022141746A1 (en) | Method for detecting anomaly in water quality and electronic device | |
| CN111176953B (en) | Abnormality detection and model training method, computer equipment and storage medium | |
| CN113672937B (en) | Block chain link point | |
| CN111046018B (en) | Multi-objective optimization-based power generation equipment operation condition library construction method and device | |
| CN102082693A (en) | Method and device for monitoring network traffic | |
| CN110083475B (en) | Abnormal data detection method and device | |
| CN108416376B (en) | SVM-based logistics man-vehicle tracking monitoring management system and method | |
| CN106027317B (en) | The Web service quality prediction system and method for trust-aware | |
| CN105871634A (en) | Method and application for detecting cluster anomalies and cluster managing system | |
| CN117705815B (en) | Printing defect detection method based on machine vision | |
| CN115242745A (en) | Short message pushing system and method based on 5G communication network | |
| CN108734226A (en) | Decision fusion method, apparatus and system | |
| CN112985440B (en) | Method, device, storage medium and program product for detecting travel track deviation | |
| CN115880101B (en) | Water conservancy data management system based on big data | |
| CN107705037A (en) | A kind of QoX quality systems system and its method of work towards contextual information processing | |
| CN106960183A (en) | A kind of image pedestrian's detection algorithm that decision tree is lifted based on gradient | |
| CN114819527B (en) | Water environment monitoring data processing method based on river growth system | |
| CN115810030A (en) | Target tracking method, device, equipment, storage medium and program product | |
| Zhang et al. | Accident Detection and Flow Prediction for Connected and Automated Transport Systems | |
| WO2022062777A1 (en) | Data management method, data management apparatus, and storage medium | |
| CN108549913A (en) | Improvement K-means clustering algorithms based on density radius | |
| CN113269238A (en) | Data stream clustering method and device based on density peak value | |
| CN112685390A (en) | Database instance management method and device and computing equipment |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181102 |