CN114997341A - Information fusion processing method and device - Google Patents

Information fusion processing method and device Download PDF

Info

Publication number
CN114997341A
CN114997341A CN202210915497.4A CN202210915497A CN114997341A CN 114997341 A CN114997341 A CN 114997341A CN 202210915497 A CN202210915497 A CN 202210915497A CN 114997341 A CN114997341 A CN 114997341A
Authority
CN
China
Prior art keywords
uncertainty
determining
system state
action
data
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
CN202210915497.4A
Other languages
Chinese (zh)
Inventor
黄安付
郭伟
吴嘉阳
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.)
Baiyang Times Beijing Technology Co ltd
Original Assignee
Baiyang Times Beijing Technology Co ltd
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 Baiyang Times Beijing Technology Co ltd filed Critical Baiyang Times Beijing Technology Co ltd
Priority to CN202210915497.4A priority Critical patent/CN114997341A/en
Publication of CN114997341A publication Critical patent/CN114997341A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
    • G01D21/00—Measuring or testing not otherwise provided for
    • G01D21/02—Measuring two or more variables by means not covered by a single other subclass
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The application discloses an information fusion processing method and device. Determining a system state according to first data by acquiring the first data, determining a first uncertainty and a second uncertainty according to the system state, and determining a reward value according to the first uncertainty and the second uncertainty; determining a first selection value according to the reward value, determining a first action according to the first selection value, and processing a fusion result corresponding to the first data according to the first action. Therefore, two uncertainties are considered respectively, the two uncertainties are measured, and the information quality is measured. And carrying out self-adaptive conflict processing through the combination of reinforcement learning and two-two uncertainty, and selecting a corresponding processing action. The self-adaptive management of the conflict evidence is realized, the accuracy of multi-sensor information fusion is improved, and the information loss is reduced.

Description

Information fusion processing method and device
Technical Field
The application relates to the technical field of combat information fusion, in particular to an information fusion processing method and device.
Background
Dempster-Shafer (evidence fusion method) is an uncertainty reasoning theory that can handle uncertain information without prior probability. Due to the characteristics of Dempster-Shafer theory, it has been widely used in military and civil fields. The Dempster combination rule is used as a classical combination rule of multi-source information fusion, and has some problems in application. Such as when the evidence to be combined is highly conflicting, it may produce an counterintuitive result.
At present, the existing method for processing the conflicting evidence focuses mainly on the original basic probability distribution, and the application of the inverse basic probability distribution is not considered in the existing method, and the inverse basic probability distribution is also an information representation mode for observing objects from a reverse side. Therefore, the data is not comprehensive and the real-time conflict processing cannot be solved, and the calculation is extremely complicated when the data amount is large.
Disclosure of Invention
In view of this, embodiments of the present application provide an information fusion processing method and apparatus, which aim to implement information fusion processing.
In a first aspect, an information fusion processing method includes:
acquiring first data, wherein the first data are basic probability distribution values acquired by a plurality of detection devices based on an identification frame, and the identification frame comprises a plurality of identification targets;
determining a system state according to the first data, wherein the system state corresponds to a fusion result;
determining a first uncertainty indicating an uncertainty of a base probability assignment and a second uncertainty indicating an uncertainty of an inverse base probability assignment based on the system state;
determining a reward value based on the first uncertainty and the second uncertainty;
determining a first selection value according to the reward value, determining a first action according to the first selection value, wherein the first selection value is used for determining a processing action corresponding to a first moment;
and processing a fusion result corresponding to the first data according to the first action.
Optionally, the determining the system state according to the first data includes:
obtaining an action set, wherein the action set is used for indicating different processing actions on the system state, and comprises a plurality of action elements;
determining a set of system states from the set of actions and the first data, the set of system states comprising a plurality of system states.
Optionally, the determining the reward value according to the first uncertainty and the second uncertainty comprises:
and acquiring a reward value corresponding to the first moment according to the magnitude relation between the first initial uncertainty and the first delayed uncertainty and between the second initial uncertainty and the second delayed uncertainty.
Optionally, the method further includes:
obtaining a set of quality values, wherein the set of quality values comprises a plurality of quality values corresponding to the identification frame;
determining a set of correlation coefficients corresponding to the identification target according to the quality value set and a first reference combination;
acquiring a first correlation coefficient, wherein the first correlation coefficient is the correlation coefficient with the largest value in the correlation coefficient set;
and determining the recognition target corresponding to the first correlation coefficient as a final decision result.
Optionally, the processing the fusion result corresponding to the first data according to the first action includes:
and in response to the first action indicating that the system state corresponding to the first data is in a conflict state, deleting the system state corresponding to the first data.
In a second aspect, an embodiment of the present application provides an information fusion processing apparatus, including:
the device comprises a first data acquisition module, a second data acquisition module and a data processing module, wherein the first data acquisition module is used for acquiring first data, and the first data is a basic probability distribution value which is acquired by a plurality of detection devices and is based on an identification frame, and the identification frame comprises a plurality of identification targets;
the system state determining module is used for determining a system state according to the first data, and the system state corresponds to the fusion result;
an uncertainty determination module to determine a first uncertainty and a second uncertainty based on the system state, the first uncertainty to indicate an uncertainty of a base probability assignment and the second uncertainty to indicate an uncertainty of an inverse base probability assignment;
a reward value determination module for determining a reward value based on the first uncertainty and the second uncertainty;
a first selection value determination module for determining a first selection value according to the reward value;
the first action determining module is used for determining a first selection value according to the reward value and determining a first action according to the first selection value, wherein the first selection value is used for determining a processing action corresponding to a first moment;
and the processing module is used for processing the fusion result corresponding to the first data according to the first action.
Optionally, the system status determining module includes:
an action set obtaining module, configured to obtain an action set, where the action set is used to indicate different processing actions on the system state, and the action set includes multiple action elements;
and the state determination execution module is used for determining a system state set according to the action set and the first data, wherein the system state set comprises a plurality of system states.
Optionally, the uncertainty determining module includes:
a time system state obtaining module, configured to obtain a first system state corresponding to the first time and a second system state corresponding to a second time, where the first time is earlier than the second time;
and the uncertainty determination execution module is used for acquiring a first initial uncertainty and a second initial uncertainty corresponding to the first system state, and a first delayed uncertainty and a second delayed uncertainty corresponding to the second system state.
Optionally, the apparatus further comprises:
the quality value acquisition module is used for acquiring a quality value set, wherein the quality value set comprises a plurality of quality values corresponding to the identification frame;
a correlation coefficient set determining module, configured to determine a correlation coefficient set corresponding to the identification target according to the quality value set and a first reference combination;
a first correlation coefficient obtaining module, configured to obtain a first correlation coefficient, where the first correlation coefficient is a correlation coefficient with a largest value in the correlation coefficient set;
and the decision result determining module is used for determining the identification target corresponding to the first correlation coefficient as a final decision result.
The embodiment of the application provides an information fusion processing method and device. In carrying out the method, two uncertainties are considered separately: uncertainty of the underlying probability assignment and uncertainty of the inverse underlying probability assignment. And measuring the uncertainty of the two parameters to measure the information quality. And carrying out self-adaptive conflict processing through combination of reinforcement learning and two-by-two uncertainty, and selecting corresponding processing actions. Therefore, the method realizes the calculation of the inverse basic probability distribution of the evidence, measures the uncertainty of the evidence and realizes the data deletion by utilizing reinforcement learning. Through the research on the original basic probability distribution and the inverse basic probability distribution, the obtained information can be more comprehensive. The innovation is that reinforcement learning and uncertainty are combined to process conflict evidence, and the contrary evidence is introduced into the reinforcement learning to realize the evaluation of information quality. Self-adaptive management of the conflict evidence is realized, the accuracy of multi-sensor information fusion is improved, and information loss is reduced.
Drawings
To illustrate the technical solutions in the present embodiment or the prior art more clearly, the drawings needed to be used in the description of the embodiment or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a flowchart of a method of information fusion processing according to an embodiment of the present application;
FIG. 2 is a flow chart of a method of information fusion processing according to an embodiment of the present disclosure;
fig. 3 is a schematic structural diagram of an apparatus for information fusion processing according to an embodiment of the present disclosure.
Detailed Description
In order to make the technical solutions of the present application better understood, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present application without making any creative effort belong to the protection scope of the present application.
As mentioned earlier, Dempster-Shafer theory has found widespread use in both military and civilian applications due to its characteristics. The Dempster combination rule is used as a classical combination rule of multi-source information fusion, and can be widely applied to a Dempster-Shafer theory of information fusion to process uncertain information without prior information. However, the inventors have studied and found that when the evidences to be combined are highly conflicting, it may lead to a counter-intuitive result. Existing methods are not sufficient to deal with conflicting evidence of real-time and online.
In order to solve the problem, an embodiment of the present application provides an information fusion processing method and apparatus. In carrying out the method, two uncertainties are considered separately: uncertainty of the underlying probability assignment and uncertainty of the inverse underlying probability assignment. And measuring the uncertainty of the two parameters to measure the information quality. And carrying out self-adaptive conflict processing through combination of reinforcement learning and two-by-two uncertainty, and selecting corresponding processing actions. Thereby, the following beneficial effects are achieved: the original basic probability distribution and the inverse basic probability distribution are considered, positive information of the evidence can be obtained from the original basic probability distribution, and negative information of the evidence can be obtained from the inverse basic probability distribution. Through the research on the original basic probability distribution and the inverse basic probability distribution, the obtained information can be more comprehensive. And introducing the contrary evidence into reinforcement learning to realize the evaluation of the information quality. Uncertainty of original evidence and inverse evidence thereof is obtained by utilizing the Duncare entropy. And then the quality of the evidence information is distinguished by the comprehensive uncertainty, thereby being beneficial to realizing the acquisition of the information. In order to realize self-adaptive online information fusion, conflict evidence is processed by combining reinforcement learning and uncertainty. In the process, a Markov decision process model is established, and a Q learning algorithm is used for solving, so that evidence fusion is realized. In general, the scheme realizes the self-adaptive management of the conflict evidence, improves the accuracy of the multi-sensor information fusion and reduces the information loss.
The method provided by the embodiment of the application can be implemented by using sensors and a processing system, assuming that five sensors detect the same target at the same time in a battlefield, and setting the identification frame to be Θ = { a, B, C }, namely that the target may be one of a (fighter), a (unmanned aerial vehicle) and a (bird). The basic probability distribution values obtained from the five sensors are respectively
Figure 123823DEST_PATH_IMAGE001
. The processing system can then perform the processing of each evidence based on the underlying probability distribution values obtained from the sensors and the settings of the method parameters (e.g., discount factor, learning rate, number of cycles).
The information fusion processing method provided by the present application is explained below by an embodiment. Referring to fig. 1, fig. 1 is a flowchart of a method of information fusion processing provided in an embodiment of the present application, including:
s101: first data is acquired.
Wherein the first data is a basic probability distribution value obtained by a detection device.
For example, in a practical application scenario, five sensors in a battlefield detect the same target at the same time, and the recognition framework can be set as
Figure 288088DEST_PATH_IMAGE002
I.e. the target may be one of a (fighter), B (drone), C (flying bird). The basic probability distribution values obtained from the five sensors are respectively
Figure 795293DEST_PATH_IMAGE001
。
Wherein, m (A) corresponds to sensor 1 data, sensor 2 data, sensor 3 data, sensor 4 data and sensor 5 data which are respectively 0.41, 0, 0.58, 0.55 and 0.60. Sensor 1 data is m 1 Wherein m (A), m (B), m (C) and m (A and C) are respectively 0.41, 0.29 and 0.30.
The data can be as shown in table 1 below.
TABLE 1 sensor detection data (BPA)
Figure 132733DEST_PATH_IMAGE003
S102: and determining the system state according to the first data.
Wherein the system state corresponds to a fusion result.
In a fusion decision system, the next state is obtained by selecting an action in the current system state. And an MDP model of the multi-sensor information fusion decision system is established.
Due to the influence of the actual environment, the multi-sensor information fusion decision-making system may have high conflict, and therefore a reasonable action strategy needs to be formulated to achieve effective processing of conflict data. In our proposed method, a set of actions
Figure 510625DEST_PATH_IMAGE004
Is defined as:
Figure DEST_PATH_IMAGE005
evidence can be through actions
Figure 376950DEST_PATH_IMAGE006
The information is retained and can be fused later. By passing
Figure 135565DEST_PATH_IMAGE007
Actions can delete high conflict evidences, and adverse effects of the conflict evidences on the fusion result are avoided. Evidence with a lower degree of conflict or less information can be acted upon
Figure 11117DEST_PATH_IMAGE008
Temporarily reserved, i.e., "pending". In the next step, a "pending" proof is manipulated. After the first round of screening all the evidence, the "to-be-processed" evidence will be processed again. Specifically, all evidence retained in the first round is fused and represented as
Figure 243515DEST_PATH_IMAGE009
. The evidence of "pending" is then reconsidered until the uncertainty of the evidence obtained by the merging is satisfied.
In reinforcement learning, when an action is taken, the state of the system will change in another state. In a fusion system, when the system behavior changes, the fusion result also changes. The current fused result can therefore be defined as the system state, i.e.,
Figure 546321DEST_PATH_IMAGE010
wherein,
Figure DEST_PATH_IMAGE011
is shown in time
Figure 293697DEST_PATH_IMAGE012
As a result of the fusion of (a),
Figure 910623DEST_PATH_IMAGE013
is at time of
Figure 794265DEST_PATH_IMAGE014
And sensor evidence of (a), and
Figure 533551DEST_PATH_IMAGE015
is shown in time
Figure 705907DEST_PATH_IMAGE014
The action taken. Based on the above analysis, the set of system states may be defined as:
Figure 923261DEST_PATH_IMAGE016
s103: a first uncertainty and a second uncertainty are determined based on the system state.
The reward is that the environment is in a certain state
Figure 130252DEST_PATH_IMAGE017
And certain actions
Figure 774860DEST_PATH_IMAGE018
Given the feedback values below, the environment in this document mainly contains sensor information and fusion results at each time instant. The system uses the reward value to determine the best operation at a time. In the method, the Deng entropy can be used for evaluating the quality of the fusion result, so that a reward function is set. The uncertainty of the original fundamental probability distribution is defined as
Figure 434511DEST_PATH_IMAGE019
. Meanwhile, the Deng entropy is used for calculating
Figure 691442DEST_PATH_IMAGE020
Is defined as negative uncertainty of
Figure 815256DEST_PATH_IMAGE021
. These two uncertainties are expressed as:
Figure 834028DEST_PATH_IMAGE022
s104: determining a reward value based on the first uncertainty and the second uncertainty.
The reward is that the environment is in a certain state
Figure 43292DEST_PATH_IMAGE017
And certain actions
Figure 540133DEST_PATH_IMAGE018
Given the feedback values below, the environment in this document mainly contains sensor information and fusion results at each time instant. The system uses the reward value to determine the best operation at a time. In the method, the Dun entropy is used for evaluating the quality of the fusion result, so that a reward function is set. The uncertainty of the original fundamental probability distribution is defined as
Figure 518453DEST_PATH_IMAGE019
. Meanwhile, the Deng entropy is also utilized to calculate
Figure 708126DEST_PATH_IMAGE020
Is defined as a negative uncertainty
Figure 404687DEST_PATH_IMAGE021
. These two uncertainties are expressed as
Figure 705218DEST_PATH_IMAGE023
Then combined with
Figure 538045DEST_PATH_IMAGE024
And
Figure DEST_PATH_IMAGE025
to judge the quality of the information. Specifically, the following cases can be classified:
case 1: if it is not
Figure 695356DEST_PATH_IMAGE026
This indicates a new state
Figure 816896DEST_PATH_IMAGE027
With less uncertainty in both positive and negative aspects, positive rewards should be awarded because adding new evidence results in a more definitive fused result.
Case 2: if it is not
Figure 717856DEST_PATH_IMAGE028
This indicates a new state
Figure 608452DEST_PATH_IMAGE027
With greater uncertainty in both positive and negative aspects, penalty rewards should be awarded because adding new evidence results in more uncertain fusion results.
Case 3: if it is not
Figure 429341DEST_PATH_IMAGE029
Or alternatively
Figure 834914DEST_PATH_IMAGE030
This indicates a new state
Figure 742827DEST_PATH_IMAGE027
Is undeterminable and is not rewarded or penalized. Thus, evidence for this case is awaiting processing.
By setting the three cases, we can adopt different strategies (i.e. delete, retain or wait for processing) for the sensor, thereby deleting the high conflict evidence and retaining the valid evidence. Based on the above analysis, the reward function herein is defined as:
Figure 550246DEST_PATH_IMAGE031
s105: a first selection value is determined based on the reward value, and a first action is determined based on the first selection value.
After modeling the MDP, the method adoptsA model-independent Q learning algorithm is used to obtain an optimal strategy. Q learning derives high quality evidence by removing the basic probability distribution of collisions. At time t, the system receives the underlying probability distribution values sent by the information sources from the different sensors and then uses an action selection strategy to select an action. Use is made here of
Figure 252623DEST_PATH_IMAGE032
Policy to select actions
Figure 145493DEST_PATH_IMAGE033
Explore new actions with probability of
Figure 325938DEST_PATH_IMAGE034
Selects the optimal action currently under consideration.
Figure 253443DEST_PATH_IMAGE032
The strategy ensures a balance between exploration and utilization of the algorithm. The specific definition is as follows:
Figure DEST_PATH_IMAGE035
wherein
Figure 923459DEST_PATH_IMAGE020
All of the optional actions are represented as,
Figure 38045DEST_PATH_IMAGE036
represent
Figure 22182DEST_PATH_IMAGE037
Function is in state
Figure 804193DEST_PATH_IMAGE017
And actions
Figure 848373DEST_PATH_IMAGE018
In (1)
Figure 217299DEST_PATH_IMAGE037
The value is obtained. The fusion system then performs the operation there and obtains a new fusion result (i.e., a new BPA). At the time of
Figure 5127DEST_PATH_IMAGE012
Measuring uncertainty of original basic probability distribution and inverse basic probability distribution by using Dun entropy, and calculating uncertainty of original basic probability distribution and inverse basic probability distribution with time
Figure 641644DEST_PATH_IMAGE038
Is compared to obtain time according to a reward function
Figure 856725DEST_PATH_IMAGE012
The prize value of. The current Q value is calculated by the following formula and stored in a Q table:
Figure 945904DEST_PATH_IMAGE039
wherein
Figure 537422DEST_PATH_IMAGE040
Is a discount factor. The fusion system selects actions according to the Q-value function, and then the system state transitions to the next state
Figure 28446DEST_PATH_IMAGE027
. As Q learning continues to explore, we update the Q value function using the following formula:
Figure 414428DEST_PATH_IMAGE041
wherein
Figure 256482DEST_PATH_IMAGE042
Is the learning rate.
S106: and processing a fusion result corresponding to the first data according to the first action.
The optimal action can then be obtained by the following formula. The system randomly selects an action with a certain probability to ensure that the algorithm has certain exploratory property. Finally, an optimal strategy is obtained:
Figure 182850DEST_PATH_IMAGE043
according to the above process, the fusion system obtains the optimal action by repeatedly calculating and updating the Q value. As a result, conflicting BPA are deleted, consistent BPA is retained, and adaptive online information processing is achieved.
In practical applications, the discounting factor (γ) = 0.9, the learning rate (α) = 0.1, and the number of cycles (M) = 100 are set according to the sensor detection data (BPA) and the corresponding method parameter set in table 1.
The results of the processing of each evidence are shown in table 2 below, giving the final result that sensor 1, sensor 3, sensor 4, sensor 5 are retained, while sensor 2 is deleted because its BPA is very different from the BPA of the other sensors and is more likely to conflict. And after the evidence body is processed, the inverse basic probability distribution value of each reserved sensor can be obtained through calculation.
TABLE 2 inverse basis probability distribution results
Figure 466064DEST_PATH_IMAGE044
⨂ means that a Dempster combination rule is utilized for fusion, the table can be subsequently applied to a decision scheme based on correlation coefficients, the correlation coefficients between each baseline BPA and the BPA obtained by combination are calculated, and corresponding propositions are determined as decision results.
The information fusion processing method provided in the embodiment of the present application is described in detail below. Referring to fig. 2, fig. 2 is another schematic flow chart of the information fusion processing method provided in the embodiment of the present application. The specific process is as follows:
s201: first data is acquired.
The first data is a basic probability distribution value acquired by the detection device.
S202: an action set is obtained.
The action set is used for indicating different processing actions on the system state, and comprises a plurality of action elements.
In the actual application process, the action set
Figure DEST_PATH_IMAGE045
Is defined as:
Figure 85264DEST_PATH_IMAGE046
wherein a is 1 ，a 2 ，a 3 And indicating actions to be reserved, deleted and processed for action elements in the action set.
S203: determining a set of system states from the set of actions and the first data.
Determining a fusion result according to the combination of each action element in the action set and the first data, defining the current fusion result as a system state, namely,
Figure 149035DEST_PATH_IMAGE010
wherein,
Figure 816777DEST_PATH_IMAGE047
is shown in time
Figure 780928DEST_PATH_IMAGE048
As a result of the fusion of (a),
Figure 102188DEST_PATH_IMAGE013
is at time of
Figure 122097DEST_PATH_IMAGE014
And sensor proof of (2), and
Figure 859109DEST_PATH_IMAGE015
is shown in time
Figure 179232DEST_PATH_IMAGE014
The action taken.
Based on the above analysis, the set of system states can be defined as:
Figure 77917DEST_PATH_IMAGE049
s204: and acquiring a first system state corresponding to the first moment and a second system state corresponding to the second moment.
Because the system state set is
Figure 116281DEST_PATH_IMAGE050
In which S is t Corresponding to the fusion result at time t, S t+1 Corresponding to the fusion result at time t + 1. In this step, the corresponding system state can therefore be determined from the set of system states as a function of the time of day value
Wherein the first time is earlier than the second time.
S205: and acquiring a first initial uncertainty and a second initial uncertainty corresponding to the first system state, and a first delayed uncertainty and a second delayed uncertainty corresponding to the second system state.
For a detailed process of determining the uncertainty according to the system state, reference is made to the foregoing embodiments, which are not described herein again.
Figure 188142DEST_PATH_IMAGE022
S206: and acquiring the reward value corresponding to the first moment according to the first initial uncertainty and the first delayed uncertainty and the magnitude relation between the second initial uncertainty and the second delayed uncertainty.
Figure DEST_PATH_IMAGE051
The process of determining the bonus value by combining the magnitude relationship between the first certainty degree and the second certainty degree is detailed in step S104, which is not described herein again.
S207: a first selection value is determined based on the reward value, and a first action is determined based on the first selection value.
The obtained bonus value is used to determine a first selection value and a first action, and for details, refer to step S105, which is not described herein.
S208: and processing a fusion result corresponding to the first data according to the first action.
And when the fusion result is processed and applied, in response to the indication that the system state corresponding to the first data is in a conflict state by the first action in an application scene, deleting the system state corresponding to the first data.
S209: a set of quality values is obtained.
Basic probability distribution
Figure 97192DEST_PATH_IMAGE052
Is fully assigned to each element in the recognition framework, i.e.
Figure 730561DEST_PATH_IMAGE053
。
S210: and determining a set of correlation coefficients corresponding to the identification target according to the quality value set and the first reference combination.
S211: a first correlation coefficient is obtained.
The correlation coefficient between each baseline BPA and the BPA combined was calculated.
In the practical application process, through calculation
Figure 928324DEST_PATH_IMAGE054
，
Figure 69455DEST_PATH_IMAGE055
,
Figure 301854DEST_PATH_IMAGE056
Correlation coefficients were obtained in combination with each baseline BPA. As shown in the following table:
TABLE 3 decision results
Figure 604659DEST_PATH_IMAGE057
S212: and determining the recognition target corresponding to the first correlation coefficient as a final decision result.
The proposition corresponding to the maximum correlation coefficient is the decision result:
Figure 555297DEST_PATH_IMAGE058
wherein
Figure 234540DEST_PATH_IMAGE059
Is the result of the final decision-making,
Figure 321445DEST_PATH_IMAGE060
(. cndot.) is the correlation coefficient.
In the practical application process, as can be seen from step S211, the proposition with the largest correlation coefficient is a, so the final decision result is a.
The foregoing provides some specific implementation manners of the information fusion processing method for the embodiments of the present application, and based on this, the present application also provides a corresponding apparatus. The device provided by the embodiment of the present application will be described in terms of functional modularity.
Referring to fig. 3, fig. 3 is a schematic structural diagram of an information fusion processing apparatus according to an embodiment of the present disclosure.
In this embodiment, the apparatus may include:
a first data obtaining module 301, configured to obtain first data, where the first data is a basic probability distribution value obtained by a plurality of detection devices based on an identification frame, and the identification frame includes a plurality of identification targets;
a system status determining module 302, configured to determine a system status according to the first data, where the system status corresponds to the fusion result;
an uncertainty determination module 303 configured to determine a first uncertainty indicating an uncertainty of the base probability assignment and a second uncertainty indicating an uncertainty of the inverse base probability assignment based on the system state;
a reward value determination module 304 for determining a reward value based on the first uncertainty and the second uncertainty;
a first selection value determining module 305 for determining a first selection value according to the reward value;
a first action determining module 306, configured to determine a first selection value according to the reward value, and determine a first action according to the first selection value, where the first selection value is used to determine a processing action corresponding to a first time;
and the processing module 307 is configured to process a fusion result corresponding to the first data according to the first action.
The system state determination module includes:
an action set acquisition module, configured to acquire an action set, where the action set is used to indicate different processing actions on the system state, and the action set includes a plurality of action elements;
and the state determination execution module is used for determining a system state set according to the action set and the first data, wherein the system state set comprises a plurality of system states.
The uncertainty determination module comprises:
a time system state obtaining module, configured to obtain a first system state corresponding to the first time and a second system state corresponding to a second time, where the first time is earlier than the second time;
and the uncertainty determination execution module is used for acquiring a first initial uncertainty and a second initial uncertainty corresponding to the first system state, and a first delayed uncertainty and a second delayed uncertainty corresponding to the second system state.
The device further comprises:
the quality value acquisition module is used for acquiring a quality value set, wherein the quality value set comprises a plurality of quality values corresponding to the identification frame;
a correlation coefficient set determining module, configured to determine a correlation coefficient set corresponding to the identification target according to the quality value set and a first reference combination;
a first correlation coefficient obtaining module, configured to obtain a first correlation coefficient, where the first correlation coefficient is a correlation coefficient with a largest value in the correlation coefficient set;
and the decision result determining module is used for determining the identification target corresponding to the first correlation coefficient as a final decision result.
The information fusion processing method and device provided by the present application are described in detail above. The embodiments are described in a progressive manner in the specification, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description. It should be noted that, for those skilled in the art, it is possible to make several improvements and modifications to the present application without departing from the principle of the present application, and such improvements and modifications also fall within the scope of the claims of the present application.
It is further noted that, in the present specification, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The above description is only exemplary of the present application and should not be taken as limiting the present application, as any modification, equivalent replacement, or improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims (10)

1. An information fusion processing method, characterized in that the method comprises:
acquiring first data, wherein the first data are basic probability distribution values acquired by a plurality of detection devices based on an identification frame, and the identification frame comprises a plurality of identification targets;
determining a system state according to the first data, wherein the system state corresponds to a fusion result;
determining a first uncertainty indicating an uncertainty of a base probability assignment and a second uncertainty indicating an uncertainty of an inverse base probability assignment based on the system state;
determining a reward value based on the first uncertainty and the second uncertainty;
determining a first selection value according to the reward value, determining a first action according to the first selection value, wherein the first selection value is used for determining a processing action corresponding to a first moment;
and processing a fusion result corresponding to the first data according to the first action.
2. The method of claim 1, wherein determining a system state from the first data comprises:
obtaining an action set, wherein the action set is used for indicating different processing actions on the system state, and comprises a plurality of action elements;
determining a set of system states from the set of actions and the first data, the set of system states comprising a plurality of system states.
3. The method of claim 2, wherein determining the first uncertainty and the second uncertainty from the system state comprises:
acquiring a first system state corresponding to the first moment and a second system state corresponding to the second moment, wherein the first moment is earlier than the second moment;
and acquiring a first initial uncertainty and a second initial uncertainty corresponding to the first system state, and a first delayed uncertainty and a second delayed uncertainty corresponding to the second system state.
4. The method of claim 3, wherein determining a reward value as a function of the first uncertainty and the second uncertainty comprises:
and acquiring a reward value corresponding to the first moment according to the magnitude relation between the first initial uncertainty and the first delayed uncertainty and between the second initial uncertainty and the second delayed uncertainty.
5. The method of claim 1, further comprising:
obtaining a set of quality values, wherein the set of quality values comprises a plurality of quality values corresponding to the identification frame;
determining a set of correlation coefficients corresponding to the identification target according to the quality value set and a first reference combination;
acquiring a first correlation coefficient, wherein the first correlation coefficient is the correlation coefficient with the largest value in the correlation coefficient set;
and determining the recognition target corresponding to the first correlation coefficient as a final decision result.
6. The method of claim 1, wherein the processing the fused result corresponding to the first data according to the first action comprises:
and in response to the first action indicating that the system state corresponding to the first data is in a conflict state, deleting the system state corresponding to the first data.
7. An information fusion processing apparatus characterized by comprising:
the device comprises a first data acquisition module, a second data acquisition module and a data processing module, wherein the first data acquisition module is used for acquiring first data, and the first data is a basic probability distribution value which is acquired by a plurality of detection devices and is based on an identification frame, and the identification frame comprises a plurality of identification targets;
the system state determining module is used for determining a system state according to the first data, and the system state corresponds to the fusion result;
an uncertainty determination module to determine a first uncertainty and a second uncertainty based on the system state, the first uncertainty to indicate an uncertainty of a base probability assignment and the second uncertainty to indicate an uncertainty of an inverse base probability assignment;
a reward value determination module for determining a reward value based on the first uncertainty and the second uncertainty;
a first selection value determination module for determining a first selection value according to the reward value;
the first action determining module is used for determining a first selection value according to the reward value and determining a first action according to the first selection value, wherein the first selection value is used for determining a processing action corresponding to a first moment;
and the processing module is used for processing the fusion result corresponding to the first data according to the first action.
8. The apparatus of claim 7, wherein the system state determination module comprises:
an action set acquisition module, configured to acquire an action set, where the action set is used to indicate different processing actions on the system state, and the action set includes a plurality of action elements;
a state determination execution module to determine a system state set from the action set and the first data, the system state set including a plurality of system states.
9. The apparatus of claim 8, wherein the uncertainty determination module comprises:
a time system state obtaining module, configured to obtain a first system state corresponding to the first time and a second system state corresponding to a second time, where the first time is earlier than the second time;
and the uncertainty determination execution module is used for acquiring a first initial uncertainty and a second initial uncertainty corresponding to the first system state, and a first delayed uncertainty and a second delayed uncertainty corresponding to the second system state.
10. The apparatus of claim 7, further comprising:
the quality value acquisition module is used for acquiring a quality value set, wherein the quality value set comprises a plurality of quality values corresponding to the identification frame;
a correlation coefficient set determining module, configured to determine a correlation coefficient set corresponding to the identification target according to the quality value set and a first reference combination;
a first correlation coefficient obtaining module, configured to obtain a first correlation coefficient, where the first correlation coefficient is a correlation coefficient with a largest value in the correlation coefficient set;
and the decision result determining module is used for determining the identification target corresponding to the first correlation coefficient as a final decision result.
CN202210915497.4A 2022-08-01 2022-08-01 Information fusion processing method and device Pending CN114997341A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210915497.4A CN114997341A (en) 2022-08-01 2022-08-01 Information fusion processing method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210915497.4A CN114997341A (en) 2022-08-01 2022-08-01 Information fusion processing method and device

Publications (1)

Publication Number Publication Date
CN114997341A true CN114997341A (en) 2022-09-02

Family

ID=83022278

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210915497.4A Pending CN114997341A (en) 2022-08-01 2022-08-01 Information fusion processing method and device

Country Status (1)

Country Link
CN (1) CN114997341A (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120233102A1 (en) * 2011-03-11 2012-09-13 Toyota Motor Engin. & Manufact. N.A.(TEMA) Apparatus and algorithmic process for an adaptive navigation policy in partially observable environments
CN107967487A (en) * 2017-11-27 2018-04-27 重庆邮电大学 A kind of colliding data fusion method based on evidence distance and uncertainty
CN113283516A (en) * 2021-06-01 2021-08-20 西北工业大学 Multi-sensor data fusion method based on reinforcement learning and D-S evidence theory

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120233102A1 (en) * 2011-03-11 2012-09-13 Toyota Motor Engin. & Manufact. N.A.(TEMA) Apparatus and algorithmic process for an adaptive navigation policy in partially observable environments
CN107967487A (en) * 2017-11-27 2018-04-27 重庆邮电大学 A kind of colliding data fusion method based on evidence distance and uncertainty
CN113283516A (en) * 2021-06-01 2021-08-20 西北工业大学 Multi-sensor data fusion method based on reinforcement learning and D-S evidence theory

Similar Documents

Publication Publication Date Title
Bertuccelli et al. Robust UAV search for environments with imprecise probability maps
CN107016464B (en) threat estimation method based on dynamic Bayesian network
CN106682502A (en) Intrusion intension recognition system and method based on hidden markov and probability inference
CN108064047B (en) Water quality sensor network optimization deployment method based on particle swarm
Pietraszek et al. The estimation of accuracy for the neural network approximation in the case of sintered metal properties
CN115617882B (en) GAN-based time sequence diagram data generation method and system with structural constraint
CN115795535A (en) Differential private federal learning method and device for providing adaptive gradient
CN116952265A (en) Factory intelligent vehicle inspection path planning method and system based on badger optimization algorithm
CN118396482A (en) Climate change-oriented dynamic monitoring method for cultivated quality
CN115293235A (en) Method for establishing risk identification model and corresponding device
Bidyuk et al. An Approach to Identifying and Filling Data Gaps in Machine Learning Procedures
CN118465724A (en) Radar target recognition method and device
CN114997341A (en) Information fusion processing method and device
CN117332335A (en) Domino effect prediction method based on information fusion
KR20200028801A (en) Learning method and learning device for variational interference using neural network and test method and test device for variational interference using the same
CN106228029B (en) Quantification problem method for solving and device based on crowdsourcing
CN113065395A (en) Radar target new class detection method based on generation countermeasure network
Bastière Methods for multisensor classification of airborne targets integrating evidence theory
US20230126695A1 (en) Ml model drift detection using modified gan
Pekaslan et al. Noise parameter estimation for non-singleton fuzzy logic systems
CN110309562B (en) Method and device for analyzing artificial thermal warming effect and storage medium
JP2020035042A (en) Data determination device, method, and program
CN112469072B (en) Node behavior monitoring duration self-adaptive adjusting method based on Bezier curve updating
Tsyganov et al. Metaheuristic algorithms for identification of the convection velocity in the convection-diffusion transport model
Hofmann A fuzzy belief-desire-intention model for agent-based image analysis

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: 20220902