CN111368871B - Multi-feature information fusion method, device and system - Google Patents
Multi-feature information fusion method, device and system Download PDFInfo
- Publication number
- CN111368871B CN111368871B CN201911308008.3A CN201911308008A CN111368871B CN 111368871 B CN111368871 B CN 111368871B CN 201911308008 A CN201911308008 A CN 201911308008A CN 111368871 B CN111368871 B CN 111368871B
- Authority
- CN
- China
- Prior art keywords
- average
- focal element
- characteristic information
- probability
- focal
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
Abstract
The invention provides a multi-feature information fusion method, a device and a system, wherein the method comprises the following steps: acquiring a plurality of feature information sets respectively output by a plurality of devices; wherein each set of characteristic information comprises a probability of at least one focal element; calculating an average collision coefficient between each characteristic information set and other characteristic information sets; if the average collision coefficient of the characteristic information set is larger than the threshold value, calculating the average support degree of the multiple devices to each focal element; for feature information sets with an average collision coefficient with other feature information sets greater than a threshold: correcting the probability of each focal element in the characteristic information set by adopting the average support degree of each focal element; and performing fusion operation on the plurality of characteristic information sets to obtain and output a fusion result. The invention carries out correction operation on the characteristic information set with larger conflict ratio, so that a fusion result which is contrary to an actual result is not generated, and the reliability of information fusion is improved.
Description
Technical Field
The present invention relates to the field of communications technologies, and in particular, to a method, an apparatus, and a system for fusing multi-feature information.
Background
With the coming of big data era, the data volume is getting bigger and bigger, and in order to increase the data processing rate, a distributed parallel processing scheme is derived, wherein the parallel processing scheme has a plurality of parallel branches.
In the distributed parallel processing scheme, each parallel branch outputs characteristic information (the characteristic information of one parallel branch can be a branch result after data processing is performed on the parallel branch), and then information fusion is performed on the characteristic information output by a plurality of parallel branches to obtain a fusion result. For example, the plurality of parallel branches may be a plurality of sensors, the sensors output the recognition results, and then the recognition results of the plurality of sensors are subjected to information fusion to obtain a fusion result.
Information fusion of a plurality of characteristic information is an important ring in big data processing. At present, the information fusion is generally carried out by adopting an evidence theory, which is also called DS (Dempster-Shafer) evidence theory. In DS evidence theory, a plurality of parallel branches are used as a plurality of evidences, and information fusion is carried out on the evidences by using a synthesis rule, so that a fusion result is obtained.
However, DS evidence theory has certain limitations: under the condition that the identification frame is provided with three target persons, namely a, b and c, and two sensors, when one person to be determined appears, the probability that the first sensor determines that the target person is a target person is 0.9, and the probability that the target person is b is 0.1; the second sensor determines that the probability of the b target person is 0.1 and the probability of the c target person is 0.9.
In this case, the recognition result of the first sensor is the target person "a", and the recognition result of the second sensor is the target person "c", which means that the collision between the two evidences is relatively large. And in the DS evidence theory, information fusion is carried out according to a synthesis rule, and an obtained fusion result is a target character b, which is obviously not in accordance with an actual scene.
That is, when the conventional DS evidence theory uses a composition rule to fuse information, when the conflict between the evidences is large, a fusion result that is contrary to the actual result is generated, and the reliability of information fusion is low.
Disclosure of Invention
In view of this, the present invention provides a method, an apparatus, and a system for fusing multi-feature information, which can improve a fusion process, so that a fusion result can be closer to an actual result, and reliability can be improved.
In order to achieve the above object, the present invention provides the following technical features:
a multi-feature information fusion method comprises the following steps:
acquiring a plurality of characteristic information sets respectively output by a plurality of devices; wherein each set of characteristic information comprises a probability of at least one focal element;
calculating an average conflict coefficient between each characteristic information set and other characteristic information sets;
if the average collision coefficient of the characteristic information set is larger than the threshold value, calculating the average support degree of the multiple devices to each focal element;
for feature information sets with an average collision coefficient with other feature information sets greater than a threshold: correcting the probability of each focal element in the characteristic information set by adopting the average support degree of each focal element;
and performing fusion operation on the plurality of characteristic information sets to obtain and output a fusion result.
Optionally, the calculating an average collision coefficient between each feature information set and another feature information set includes:
calculating the average collision coefficient between each characteristic information set and other characteristic information sets by adopting the following formula
Wherein, the first and the second end of the pipe are connected with each other,
wherein the content of the first and second substances,
wherein the content of the first and second substances,
represents m i And m j The distance between them;
wherein the content of the first and second substances,
is the average collision coefficient, wf, between the ith feature information set and other feature information i,l Is the collision coefficient between the ith characteristic information set and the ith characteristic information set, n is the number of a plurality of devices, i, l =1,2 … … n, k il The original conflict coefficient between the i characteristic information sets and the l characteristic information set is obtained;
A j j =1,2 … … M, where M is the total number of all focal elements; m is a probability function, m i (A j ) Represents the ith characteristic information concentration focal element A j The probability of (d);
is a 2 n ×2 n Of the matrix of (a).
Optionally, the calculating an average support degree of the multiple devices for each focal element includes:
calculating the average probability of each focal element of a plurality of devices;
and determining the average probability of the plurality of devices to each focal element as the average support degree of the plurality of devices to each focal element.
Optionally, the calculating an average support degree of the multiple devices for each focal element includes:
calculating each focal element A by adopting the following formula j Average support degree of (A) j )：
Wherein A is j Denotes the j-th focal element, j =1,2 … … M, M is the total number of all focal elements, S (A) j ) Represents the average support degree of the jth focal element, m is a probability function, m i (A j ) Denotes the firsti characteristic information concentration focal element A j N is the number of the plurality of devices.
Optionally, the calculating an average support degree of the multiple devices for each focal element includes:
calculating the average probability of each focal element of a plurality of devices;
calculating the information energy ratio of each focal element in all the focal elements based on the average probability of each focal element;
and determining the information energy ratio of each focal element as the average support degree of a plurality of devices to each focal element.
Optionally, the calculating an average support degree of the multiple devices for each focal element includes:
calculating each focal element A by adopting the following formula j Average degree of support of H (A) j )：
Wherein the content of the first and second substances,
wherein A is j Denotes the j-th focal element, j =1,2 … … M, M is the total number of all focal elements, H (A) j ) Denotes the average support of the j-th focal element, S (A) j ) Representing the average probability of a plurality of devices for the jth focal element; m is a probability function, m i (A j ) Represents the ith characteristic information concentration focal element A j N is the number of the plurality of devices.
Optionally, the modifying the probability of each focal element in the feature information set by using the average support degree of each focal element includes:
calculating the product of the probability of each focal element in the characteristic information set and the average support degree of each focal element to obtain the correction probability of each focal element in the characteristic information set;
and calculating the probability of the characteristic information set that each focal element forms a complete set based on the correction probability of each focal element in the characteristic information set.
Optionally, the performing the fusion operation on the plurality of feature information sets to obtain and output a fusion result includes performing the fusion operation on the plurality of feature information sets according to an update synthesis formula to obtain the fusion result;
and the updating synthesis formula comprises the step of increasing the average support degree and the product of the value and the conflict coefficient of each focal element on the basis of the original synthesis formula.
A multi-feature information fusion apparatus comprising:
an acquisition unit configured to acquire a plurality of feature information sets output by a plurality of devices, respectively; wherein each set of characteristic information comprises a probability of at least one focal element;
a conflict coefficient determining unit for calculating an average conflict coefficient between each feature information set and other feature information sets;
the average support degree calculating unit is used for calculating the average support degree of the plurality of devices to each focal element if the average collision coefficient of the characteristic information set is larger than a threshold value;
a correction unit configured to, for a feature information set whose average collision coefficient with other feature information sets is larger than a threshold: correcting the probability of each focal element in the characteristic information set by adopting the average support degree of each focal element;
and the fusion unit is used for performing fusion operation on the plurality of characteristic information sets to obtain and output a fusion result.
A multi-feature information fusion system, comprising:
a plurality of devices and an information fusion device;
the plurality of devices are used for respectively outputting a plurality of feature information sets; wherein each set of characteristic information comprises a probability of at least one focal element;
the information fusion device is used for acquiring a plurality of characteristic information sets output by a plurality of devices respectively and calculating an average collision coefficient between each characteristic information set and other characteristic information sets; if the average collision coefficient of the characteristic information set is larger than the threshold value, calculating the average support degree of the multiple devices to each focal element; for feature information sets with an average collision coefficient with other feature information sets greater than a threshold: correcting the probability of each focal element in the characteristic information set by adopting the average support degree of each focal element; and performing fusion operation on the plurality of characteristic information sets to obtain and output a fusion result.
Through the technical means, the following beneficial effects can be realized:
the method calculates the average support degree of the plurality of devices to each focal element, wherein the average support degree of each focal element is obtained by comprehensively considering the probabilities of the plurality of devices, so that the average support degree of each focal element can reflect the actual probability of each focal element to a certain extent.
The invention is used for the characteristic information sets with the average conflict coefficients larger than the threshold value with other characteristic information sets: and correcting the probability of each focal element in the characteristic information set by adopting the average support degree of each focal element. Therefore, the probability of each focal element in the characteristic information set can be closer to the actual probability.
And after the correction operation is carried out on the feature information set with larger conflict ratio, the information fusion operation is carried out, and finally, a fusion result is obtained and output. Since the correction operation is performed on the feature information set with a large conflict ratio, a fusion result which is contrary to an actual result is not generated, and the reliability of information fusion is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic structural diagram of a multi-feature information fusion system disclosed in an embodiment of the present invention;
FIG. 2 is a flowchart of a first embodiment of a multi-feature information fusion method disclosed in the present invention;
FIG. 3 is a flowchart of a second embodiment of a multi-feature information fusion method disclosed in the embodiments of the present invention;
fig. 4 is a schematic structural diagram of a multi-feature information fusion apparatus according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without making any creative effort based on the embodiments in the present invention, belong to the protection scope of the present invention.
To facilitate the understanding of the system structure of the present invention by those skilled in the art, referring to fig. 1, the present invention provides a multi-feature information fusion system, comprising:
a plurality of devices 100 and an information fusion device 200.
The plurality of devices 100 for respectively outputting a plurality of feature information sets; wherein each set of characteristic information comprises a probability of at least one focal element.
The information fusion device 200 is configured to obtain a plurality of feature information sets output by a plurality of devices, and calculate an average collision coefficient between each feature information set and another feature information set; calculating the average support degree of the plurality of devices for each focal element; for feature information sets with an average collision coefficient with other feature information sets greater than a threshold: correcting the probability of each focal element in the characteristic information set by adopting the average support degree of each focal element; and performing fusion operation on the plurality of characteristic information sets to obtain and output a fusion result.
For details of a specific implementation process of the information fusion device, the following embodiments are described, and details are not repeated here.
The multiple devices provided by the invention are multiple branches in a distributed parallel processing scheme, and the information fusion device is a device for information fusion. For example, the plurality of devices may be a plurality of sensors, and the information fusion device may be a processing device such as a server or a processor.
Since the information fusion can be applied to different practical application scenarios, the specific implementation forms of the multiple devices and the information fusion device can be determined according to the practical application scenarios, which is not limited in the present invention.
Before introducing the multi-feature information fusion method, the basic concept of DS evidence theory is first introduced for the understanding of those skilled in the art. There are some definitions specified in DS evidence theory, which are described in detail below.
Definition 1: let Θ be a finite and complete sample space, and let it be N elements, all of which are mutually exclusive, and then be called the recognition framework. Can be expressed as:
Θ＝{θ 1 ，θ 2 ，……，θ N }
2 Θ the representation recognizes a set of all subsets of the framework Θ, referred to as the power set of Θ. Can be expressed as:
2 Θ ＝{Φ,{θ 1 },{θ 2 },{θ 3 },…,{θ n },{θ 1 ∪θ 2 },{θ 1 ∪θ 3 },…,Θ}
when there are N elements in Θ,2 Θ Therein is 2 N And (4) each element.
Definition 2: for the recognition framework Θ, we call the mapping m:2 Θ →[0,1]A Basic Probability assignment BPA (Basic Probability assignment) function, also called m, is assigned to identify one Basic confidence on the space Θ, and the function m satisfies:
m(Φ)＝0
where m (A) is referred to as the basic probability distribution of A, and m (A) represents the confidence of evidence m on A. The basic probability assignment for the empty set is 0.
When m (A) > 0 is satisfied and
when called A is the pyrogen of BPA. The number of identification frame elements contained in a focal element is referred to as the base of the focal element. When 2 is in Θ When the subset A contains only one element (e.g. { theta.) 1 }) called single element focal. For the same reason, 2 Θ When two elements are contained in the subset A of (e.g., { θ ] 1 ∪θ 2 }) is called a two-element focal element.
To facilitate understanding of the definition of DS evidence theory above, a practical application scenario is used for illustration. Suppose that the practical application scene is access control system, access control system has a plurality of sensors and is equivalent to a plurality of equipment, and access control system has that the treater is equivalent to information fusion equipment, just can open the entrance guard to the inside personnel of discernment, otherwise does not open the entrance guard.
In this application scenario, the information fusion device stores a plurality of internal staff members, that is, a plurality of elements in advance, and the plurality of internal staff members form an identification framework Θ = { θ = { (θ) } 1 ，θ 2 ，…θ i …，θ n In which θ i Representing an internal personnel element.
In practical scenarios, it may happen that two or several insiders are present at the same time, so there is a combination of two or more inside-person elements, e.g. { θ } 1 ∪θ 2 Or { theta } or 1 ∪θ 3 }. For convenience of reference, a single internal personnel element, or a union of two or more internal personnel elements, is referred to as a focal element.
The sensors can be respectively installed at the top of the entrance door, and under the condition that the person to be determined intends to enter, the sensors can respectively identify the person to be determined, calculate the probability of the person to be determined as the focal element by adopting a probability distribution function, and output the probability of each focal element. The probabilities of the focal elements output by one sensor constitute a feature information set.
The invention provides a first embodiment of a multi-feature information fusion method, which is applied to information fusion equipment shown in fig. 1. Referring to fig. 2, the method comprises the steps of:
step S201: acquiring a plurality of feature information sets respectively output by a plurality of devices; wherein each set of characteristic information comprises a probability of at least one focal element.
Assuming that there are n devices and M focal elements, one device can output a feature information set, and a feature information set includes the probability of multiple focal elements, then the ith device M i The j focal element A in the characteristic information set j Has a probability of m i (A j ). Wherein i =1,2 … … n, j =1,2 … … M, and M is the total number of all focal elements.
Step S202: and calculating the average collision coefficient between each characteristic information set and other characteristic information sets.
Calculating the average collision coefficient between each feature information set and other feature information sets, the average collision coefficient between each feature information set and other feature information sets can be calculated by using the following formula
Wherein the content of the first and second substances,
wherein the content of the first and second substances,
wherein the content of the first and second substances,
is the average collision coefficient, wf, between the ith feature information set and other feature information i,l Is the conflict coefficient between the ith characteristic information set and the ith characteristic information set, n is the number of a plurality of devices, i, l =1,2 … … n, k il Is the original collision coefficient between the i characteristic information sets and the l characteristic information set.
A j Represents the j-th focal element, j =1,2 … … M, M is the total number of all focal elements; m is a probability function, m i (A j ) Represents the ith characteristic information concentration focal element A j The probability of (d);
represents m i And m j Is greater than or equal to>
Is a 2 n ×2 n Of the matrix of (a).
Step S203: and if the average collision coefficient of the characteristic information set is larger than the threshold value, calculating the average support degree of the plurality of devices to each focal element.
In order to correct the feature information set with a large conflict ratio, and the conflict between the corrected feature information set and other feature information sets is small, the corrected feature information set can be close to the actual situation, so a more appropriate correction mode is needed.
For this purpose, the present embodiment proposes to calculate an average support degree of each focal element by the multiple devices, where the average support degree of each focal element is obtained by comprehensively considering the probabilities of the multiple devices, so that the average support degree of each focal element can reflect the actual probability of each focal element to some extent, and the conflict between the modified feature information set and other feature information sets can be relatively small.
The invention provides two implementation modes for calculating the average support degree of each target focal element:
the first implementation mode comprises the following steps: and taking the average probability as the average support degree.
S1: calculating the average probability of each focal element of a plurality of devices;
s2: and determining the average probability of the plurality of devices to each focal element as the average support degree of the plurality of devices to each focal element.
It can be understood that the probability that a device outputs a focal element is the support degree of the device for the focal element, a high probability indicates that the device has a high probability of confirming the focal element, that is, the support degree of the focal element is high, and a low probability indicates that the device has a low probability of confirming the focal element, that is, the support degree of the focal element is low.
Therefore, in the first implementation manner, the average probability of each focal element by the multiple devices is determined as the average support degree of each focal element by the multiple devices.
In a first implementation, the following formula can be used to calculate each focal element A j Average support degree of (A) j )：
Wherein, A j Denotes the j-th focal element, S (A) j ) Represents the average support degree of the jth focal element, m is a probability function, m i (A j ) Represents the ith characteristic information concentration focal element A j N is the number of the plurality of devices.
The second implementation mode comprises the following steps: and determining the average information energy ratio as the average support degree of the plurality of devices to each focal element.
S1: the average probability of multiple devices for each focal element is calculated.
S2: and calculating the information energy ratio of each focal element in all the focal elements based on the average probability of each focal element.
And calculating the information energy of each focal element based on the average probability of each focal element, calculating the information energy sum of each focal element, and then calculating the ratio of the information energy of each focal element to the information energy sum of each focal element, namely the information energy ratio of each focal element in all the focal elements.
S3: and determining the information energy ratio of each focal element as the average support degree of a plurality of devices to each focal element.
In the first way, the average probability of each focal element by the multiple devices can be used as the average support degree of each focal element by the multiple devices, and in order to increase the difference between the average support degrees of each focal element and make the differentiation more obvious, the information energy of each focal element is calculated based on the average probability of each focal element.
The square of the average probability of one focal element can be used as the information energy of the focal element, and the information energy is the square of the average probability, so that the difference between the average support degrees of the focal elements is increased, and the differentiation is more obvious. Therefore, the correction probability after the subsequent correction operation is closer to the actual scene.
In a second implementation, the following formula can be used to calculate each focal element A j Average degree of support of H (A) j )：
Wherein the content of the first and second substances,
wherein A is j Denotes the j-th focal element, H (A) j ) Denotes the average support of the j-th focal element, S (A) j ) Representing the average probability of a plurality of devices for the jth focal element; m is a probability function, m i (A j ) Represents the ith characteristic information concentration focal element A j N is the number of the plurality of devices.
Step S204: for feature information sets with an average collision coefficient with other feature information sets greater than a threshold: and correcting the probability of each focal element in the characteristic information set by adopting the average support degree of each focal element.
The information fusion device stores a threshold value indicating that the conflict is large in advance, and when the average conflict coefficient between one characteristic information set and other characteristic information sets is larger than the threshold value, the information fusion device indicates that the conflict between the characteristic information and the other characteristic information sets is large. This characteristic information set may be due to noise interference of the device or a failure of the device itself, and needs to be corrected in order to avoid a subsequent fusion result which is contrary to the practical situation.
It is understood that when the average collision coefficient between one feature information set and the other feature information sets is smaller than the threshold, it indicates that the collision between the feature information and the other feature information sets is not large. In order to ensure the authenticity and practicability of the feature information set, the feature information set does not need to be corrected.
In this step, the average support degree of each focal element can be used to perform correction operation on the probability of each focal element in the feature information set, and the probability of each focal element in the feature information set forming a complete set can be calculated based on the correction probability of each focal element in the feature information set.
That is, in this step, not only the probability of each focal element is revised again, but also the expansion operation is performed on the basis of each Jiao Yuanji, so that the probability of the complete set is expanded.
Corresponding to two implementations in step S203, this step also provides two implementations:
the first implementation mode comprises the following steps: the average support S (A) determined in step S203 using the first implementation j ) And carrying out correction operation.
And calculating the correction probability and the corpus probability of each focal element by adopting the following formulas:
wherein m is i '(A j ) Is m i (A j ) Correction probability of m i ' (Θ) is the extended corpus probability.
The second implementation mode comprises the following steps: the average support H (A) determined by the second implementation in step S203 j ) And carrying out correction operation.
And calculating the correction probability and the corpus probability of each focal element by adopting the following formulas:
through the first implementation manner or the second implementation manner, the feature information set with a large conflict can be corrected, and after the correction operation, the conflict between the feature information set and other feature information sets can be reduced.
Step S205: and performing fusion operation on the plurality of characteristic information sets to obtain and output a fusion result.
Through the technical means, the following beneficial effects can be realized:
the method calculates the average support degree of the plurality of devices to each focal element, wherein the average support degree of each focal element is obtained by comprehensively considering the probabilities of the plurality of devices, so that the average support degree of each focal element can reflect the actual probability of each focal element to a certain extent.
The invention is used for the characteristic information sets with the average conflict coefficient larger than the threshold value with other characteristic information sets: and correcting the probability of each focal element in the characteristic information set by adopting the average support degree of each focal element. Therefore, the probability of each focal element in the characteristic information set can be closer to the actual probability.
And after the correction operation is carried out on the feature information set with larger conflict ratio, the information fusion operation is carried out, and finally, a fusion result is obtained and output. Since the correction operation is performed on the feature information set with a large conflict ratio, a fusion result which is contrary to an actual result is not generated, and the reliability of information fusion is improved.
In order to further improve the reliability, the invention also corrects the synthesis rule of the DS evidence theory.
The invention provides a second embodiment of a multi-feature information fusion method, which is applied to the information fusion device shown in fig. 1. Referring to fig. 3, the method comprises the steps of:
step S301: acquiring a plurality of characteristic information sets respectively output by a plurality of devices; wherein each set of characteristic information comprises a probability of at least one focal element.
Step S302: and calculating the average collision coefficient between each characteristic information set and other characteristic information sets.
Step S303: and if the average collision coefficient of the characteristic information set is larger than the threshold value, calculating the average support degree of the plurality of devices to each focal element.
Step S304: for feature information sets with an average collision coefficient with other feature information sets greater than a threshold: and correcting the probability of each focal element in the characteristic information set by adopting the average support degree of each focal element.
The execution process of steps S301 to S304 is detailed in the execution process of steps S201 to S204, and is not described herein again.
Step S305: and recalculating the average support degree of the plurality of devices for each focal element.
After the correction operation is carried out on the characteristic information set of which the average collision coefficient with other characteristic information sets is larger than the threshold value, the probability of each focal element in the characteristic information set of which the average collision coefficient is larger than the threshold value is changed, and the average support degree of each focal element by a plurality of devices is recalculated for the reason.
In a first implementation of step S203, recalculating the average support of multiple devices for each focal element may be represented as S' (a) j )。
In a second implementation manner of step S203, recalculating the average support degree of multiple devices for each focal element may be represented as H' (a) j )。
Step S306: and performing fusion operation on the plurality of characteristic information sets to obtain and output a fusion result.
In the embodiment, a synthesis formula of the DS evidence theory is updated to obtain an updated synthesis formula, and fusion operation is performed on a plurality of feature information sets according to the updated synthesis formula to obtain a fusion result; wherein, updating the synthesis formula comprises increasing the product of the sum of the average support degree of each focal element and the conflict coefficient on the basis of the original synthesis formula.
Corresponding to step S305, this step provides two implementations:
the first implementation mode comprises the following steps: the average support degree S' (A) of each focal element determined in the first implementation manner in step S305 is adopted j ). The update synthesis mode can be expressed as:
wherein the content of the first and second substances,
where m (a) is the fusion result, S' (a) is the sum of the average support degrees of the respective focal elements, and k is the collision coefficient between the respective feature information sets.
The second implementation mode comprises the following steps: the average support degree H' (a) of each focal element determined in the second implementation manner in step S305 is adopted j )。
Wherein the content of the first and second substances,
/>
where m (a) is the fusion result, H' (a) is the sum of the average support degrees of the respective focal elements, and k is the collision coefficient between the respective feature information sets.
Because the average support degree of each focal element and the probability of each corrected focal element are considered again in the updating and synthesizing formula, the updating and synthesizing formula can be considered more comprehensively, and the final information fusion result is more accurate.
Referring to fig. 4, the present invention provides a multi-feature information fusion apparatus, including:
an acquisition unit 41 configured to acquire a plurality of feature information sets output by a plurality of devices, respectively; wherein each set of characteristic information comprises a probability of at least one focal element;
a conflict coefficient determining unit 42 for calculating an average conflict coefficient between each feature information set and other feature information sets;
a calculate average support degree unit 43, configured to calculate an average support degree of each focal element by the multiple devices if an average collision coefficient of a feature information set is greater than a threshold;
a modification unit 44, configured to, for a feature information set whose average collision coefficient with other feature information sets is greater than a threshold: correcting the probability of each focal element in the characteristic information set by adopting the average support degree of each focal element;
and a fusion unit 45, configured to perform fusion operation on the multiple feature information sets to obtain and output a fusion result.
The detailed implementation of each unit of the multi-feature information fusion device can refer to the embodiments shown in fig. 2 and fig. 3, and is not described herein again.
Through the technical means, the following beneficial effects can be realized:
the method calculates the average support degree of the plurality of devices to each focal element, wherein the average support degree of each focal element is obtained by comprehensively considering the probabilities of the plurality of devices, so that the average support degree of each focal element can reflect the actual probability of each focal element to a certain extent.
The invention is used for the characteristic information sets with the average conflict coefficient larger than the threshold value with other characteristic information sets: and correcting the probability of each focal element in the characteristic information set by adopting the average support degree of each focal element. This can make the probability of each focal element in the feature information set closer to the actual probability.
And after the correction operation is carried out on the feature information set with larger conflict ratio, the information fusion operation is carried out, and finally, a fusion result is obtained and output. Since the correction operation is performed on the feature information set with a large conflict ratio, a fusion result which is contrary to an actual result is not generated, and the reliability of information fusion is improved.
The functions described in the method of the present embodiment, if implemented in the form of software functional units and sold or used as independent products, may be stored in a storage medium readable by a computing device. Based on such understanding, part of the contribution of the embodiments of the present invention to the prior art or part of the technical solution may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computing device (which may be a personal computer, a server, a mobile computing device, a network device, or the like) to execute all or part of the steps of the method described in the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The embodiments are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same or similar parts among the embodiments are referred to each other.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A multi-feature information fusion method is characterized by comprising the following steps:
acquiring a plurality of characteristic information sets respectively output by a plurality of devices; wherein each set of characteristic information comprises a probability of at least one focal element;
calculating an average conflict coefficient between each characteristic information set and other characteristic information sets;
if the average conflict coefficient of the characteristic information set is larger than the threshold value, calculating the average support degree of the plurality of devices to each focal element;
for feature information sets with an average collision coefficient with other feature information sets greater than a threshold: correcting the probability of each focal element in the characteristic information set by adopting the average support degree of each focal element;
and performing fusion operation on the plurality of characteristic information sets to obtain and output a fusion result.
2. The method of claim 1, wherein calculating an average collision coefficient between each set of feature information and other sets of feature information comprises:
calculating the average collision coefficient between each characteristic information set and other characteristic information sets by adopting the following formula
Wherein the content of the first and second substances,
wherein the content of the first and second substances,
wherein the content of the first and second substances,
represents m i And m j The distance between them;
wherein the content of the first and second substances,
is the average collision coefficient, wf, between the ith feature information set and other feature information i,l Is the collision coefficient between the ith characteristic information set and the ith characteristic information set, n is the number of a plurality of devices, i, l =1,2 … … n, k il The original conflict coefficient between the i characteristic information sets and the l characteristic information set is obtained;
A j j =1,2 … … M, where M is the total number of all focal elements; m is a probability function, m i (A j ) Represents the ith characteristic information concentration focal element A j The probability of (d);
is one 2 n ×2 n Of the matrix of (a).
3. The method of claim 1, wherein calculating the average degree of support for each focal element by the plurality of devices comprises:
calculating the average probability of each focal element of a plurality of devices;
and determining the average probability of each focal element by the multiple devices as the average support degree of each focal element by the multiple devices.
4. The method of claim 3, wherein calculating the average degree of support for each focal element by the plurality of devices comprises:
calculating each focal element A by adopting the following formula j Average support degree of S (A) j )：
Wherein, A j Denotes the j-th focal element, j =1,2 … … M, M is the total number of all focal elements, S (A) j ) Represents the average support degree of the jth focal element, m is a probability function, m i (A j ) Represents the ith characteristic information concentration focal element A j N is the number of the plurality of devices.
5. The method of claim 1, wherein calculating an average degree of support for each focal element by a plurality of devices comprises:
calculating the average probability of each focal element of a plurality of devices;
calculating the information energy ratio of each focal element in all the focal elements based on the average probability of each focal element;
and determining the information energy ratio of each focal element as the average support degree of the plurality of devices to each focal element.
6. The method of claim 5, wherein calculating an average degree of support for each focal element by a plurality of devices comprises:
calculating each focal element A by adopting the following formula j Average degree of support of H (A) j )：
Wherein the content of the first and second substances,
wherein, A j Denotes the j-th focal element, j =1,2 … … M, M is the total number of all focal elements, H (A) j ) Represents the average support of the j-th focal element, S (A) j ) Representing the average probability of a plurality of devices for the jth focal element; m is a probability function, m i (A j ) Represents the ith characteristic information concentration focal element A j N is the number of the plurality of devices.
7. The method of claim 4, wherein the modifying the probability of each focal element in the feature information set using the average support of each focal element comprises:
calculating the product of the probability of each focal element in the characteristic information set and the average support degree of each focal element to obtain the correction probability of each focal element in the characteristic information set;
and calculating the probability of the characteristic information set that each focal element forms a complete set based on the correction probability of each focal element in the characteristic information set.
8. The method of claim 1, wherein performing the fusion operation on the plurality of feature information sets to obtain and output a fusion result comprises performing the fusion operation on the plurality of feature information sets according to an update synthesis formula to obtain a fusion result;
and the updating synthesis formula comprises the step of increasing the average support degree and the product of the value and the conflict coefficient of each focal element on the basis of the original synthesis formula.
9. A multi-feature information fusion apparatus, comprising:
an acquisition unit configured to acquire a plurality of feature information sets output by a plurality of devices, respectively; wherein each set of characteristic information comprises a probability of at least one focal element;
a conflict coefficient determining unit for calculating an average conflict coefficient between each feature information set and other feature information sets;
the average support degree calculating unit is used for calculating the average support degree of the plurality of devices to each focal element if the average collision coefficient of the characteristic information set is larger than a threshold value;
a correction unit configured to, for a feature information set whose average collision coefficient with other feature information sets is larger than a threshold: correcting the probability of each focal element in the characteristic information set by adopting the average support degree of each focal element;
and the fusion unit is used for performing fusion operation on the plurality of characteristic information sets to obtain and output a fusion result.
10. A multi-feature information fusion system, comprising:
a plurality of devices and an information fusion device;
the plurality of devices are used for respectively outputting a plurality of characteristic information sets; wherein each set of characteristic information comprises a probability of at least one focal element;
the information fusion device is used for acquiring a plurality of characteristic information sets output by a plurality of devices respectively and calculating an average collision coefficient between each characteristic information set and other characteristic information sets; if the average collision coefficient of the characteristic information set is larger than the threshold value, calculating the average support degree of the multiple devices to each focal element; for feature information sets with an average collision coefficient with other feature information sets greater than a threshold: correcting the probability of each focal element in the characteristic information set by adopting the average support degree of each focal element; and performing fusion operation on the plurality of characteristic information sets to obtain and output a fusion result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911308008.3A CN111368871B (en) | 2019-12-18 | 2019-12-18 | Multi-feature information fusion method, device and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911308008.3A CN111368871B (en) | 2019-12-18 | 2019-12-18 | Multi-feature information fusion method, device and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111368871A CN111368871A (en) | 2020-07-03 |
| CN111368871B true CN111368871B (en) | 2023-04-07 |
Family
ID=71206328
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911308008.3A Active CN111368871B (en) | 2019-12-18 | 2019-12-18 | Multi-feature information fusion method, device and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111368871B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106650785A (en) * | 2016-11-09 | 2017-05-10 | 河南大学 | Weighted evidence fusion method based on evidence classification and conflict measurement |
| WO2017198198A1 (en) * | 2016-05-19 | 2017-11-23 | 中兴通讯股份有限公司 | Software compiling method and device |
| CN110033028A (en) * | 2019-03-19 | 2019-07-19 | 河南大学 | Conflicting evidence fusion method based on arithmetic average approach degree |
-
2019
- 2019-12-18 CN CN201911308008.3A patent/CN111368871B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017198198A1 (en) * | 2016-05-19 | 2017-11-23 | 中兴通讯股份有限公司 | Software compiling method and device |
| CN106650785A (en) * | 2016-11-09 | 2017-05-10 | 河南大学 | Weighted evidence fusion method based on evidence classification and conflict measurement |
| CN110033028A (en) * | 2019-03-19 | 2019-07-19 | 河南大学 | Conflicting evidence fusion method based on arithmetic average approach degree |
Non-Patent Citations (2)
| Title |
|---|
| 一种新的基于证据理论的合成公式;孙全等;《电子学报》(第08期);全文 * |
| 一种有效地解决冲突证据的组合方法;胡丽芳等;《舰船电子工程》(第01期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111368871A (en) | 2020-07-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2019173734A1 (en) | Systems and methods for providing machine learning model evaluation by using decomposition | |
| CN109472419B (en) | Method and device for establishing warning condition prediction model based on time and space and storage medium | |
| CN110033028B (en) | Conflict evidence fusion method based on arithmetic mean closeness | |
| CN111079186B (en) | Data analysis method, device, equipment and storage medium | |
| CN109063921B (en) | Optimization processing method and device for client risk early warning, computer equipment and medium | |
| CN111046027A (en) | Missing value filling method and device for time series data | |
| US11238539B1 (en) | Intelligent routing control | |
| CN113222942A (en) | Training method of multi-label classification model and method for predicting labels | |
| CN113379301A (en) | Method, device and equipment for classifying users through decision tree model | |
| CN112288025A (en) | Abnormal case identification method, device and equipment based on tree structure and storage medium | |
| CN113807940B (en) | Information processing and fraud recognition method, device, equipment and storage medium | |
| CN114913487A (en) | Target recognition detection method based on multi-modal learning and related components | |
| CN114817933A (en) | Method and device for evaluating robustness of business prediction model and computing equipment | |
| CN111368871B (en) | Multi-feature information fusion method, device and system | |
| CN108108395B (en) | Mutual-aid navigation positioning data credibility evaluation method and system | |
| CN111694936A (en) | Method and device for identifying AI intelligent interview, computer equipment and storage medium | |
| CN115277250A (en) | Vehicle-end attack path identification method, equipment and storage medium | |
| CN115099875A (en) | Data classification method based on decision tree model and related equipment | |
| CN111291196B (en) | Knowledge graph perfecting method and device, and data processing method and device | |
| CN112651494A (en) | Training method of neural network for cable anti-winding early warning | |
| KR102491451B1 (en) | Apparatus for generating signature that reflects the similarity of the malware detection classification system based on deep neural networks, method therefor, and computer recordable medium storing program to perform the method | |
| CN116170229B (en) | Network security detection method, device, server and computer readable storage medium | |
| CN117057929B (en) | Abnormal user behavior detection method, device, equipment and storage medium | |
| US11461701B2 (en) | Method and apparatus for preventing model theft during model-based classification | |
| CN115827876B (en) | Method and device for determining unlabeled text and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Chen Yingchun Inventor after: Yu Daojie Inventor after: Li Ou Inventor after: Liu Guangyi Inventor after: Dong Fang Inventor after: Xue Jingjing Inventor after: Ran Xiaomin Inventor after: Mo Youquan Inventor after: Zhang Jing Inventor after: Wang Xiaomei Inventor before: Chen Yingchun Inventor before: Yu Daojie Inventor before: Li Ou Inventor before: Liu Guangyi Inventor before: Dong Fang Inventor before: Xue Jingjing Inventor before: Ran Xiaomin Inventor before: Mo Youquan Inventor before: Zhang Jing Inventor before: Wang Xiaomei |
|
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |