CN108052976A - A kind of multi-band image fusion identification method - Google Patents
A kind of multi-band image fusion identification method Download PDFInfo
- Publication number
- CN108052976A CN108052976A CN201711332225.7A CN201711332225A CN108052976A CN 108052976 A CN108052976 A CN 108052976A CN 201711332225 A CN201711332225 A CN 201711332225A CN 108052976 A CN108052976 A CN 108052976A
- Authority
- CN
- China
- Prior art keywords
- wave infrared
- probability distribution
- distribution function
- medium
- infrared image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000004927 fusion Effects 0.000 title claims abstract description 56
- 238000000034 method Methods 0.000 title claims abstract description 49
- 238000012549 training Methods 0.000 claims abstract description 48
- 238000002156 mixing Methods 0.000 claims abstract description 15
- 238000012545 processing Methods 0.000 claims abstract description 5
- 238000005315 distribution function Methods 0.000 claims description 119
- 230000008569 process Effects 0.000 claims description 26
- 230000004069 differentiation Effects 0.000 claims description 20
- 230000015572 biosynthetic process Effects 0.000 claims description 18
- 238000003786 synthesis reaction Methods 0.000 claims description 18
- 238000004458 analytical method Methods 0.000 claims description 9
- 238000001514 detection method Methods 0.000 claims description 8
- 239000000203 mixture Substances 0.000 claims description 7
- 238000000605 extraction Methods 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 230000008901 benefit Effects 0.000 abstract description 11
- 238000005516 engineering process Methods 0.000 abstract description 3
- 230000010354 integration Effects 0.000 abstract description 2
- 239000013589 supplement Substances 0.000 abstract description 2
- 238000003384 imaging method Methods 0.000 description 3
- 238000000701 chemical imaging Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005192 partition Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000003331 infrared imaging Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/251—Fusion techniques of input or preprocessed data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10048—Infrared image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of multi-band image fusion identification methods, medium-wave infrared image and LONG WAVE INFRARED image are merged, for supplementary training sample, data processing is carried out on the basis of training sample after supplement, therefore classifying and dividing can be carried out on the visible images with abundant details and color information and higher resolution ratio and using medium-wave infrared image and the respective advantage of LONG WAVE INFRARED image, by the data fusion of respective advantage maximum, therefore, the loss of information can maximumlly be reduced.The present invention builds 3 kinds of graders before this, and the blending image of medium-wave infrared image and LONG WAVE INFRARED image is as supplementary training sample, since blending image is that the feature utilized shifts to an earlier date integration technology, so that original single training sample is provided with the outstanding feature of the respective advantage of medium-wave infrared image and LONG WAVE INFRARED image, the probability of the respective advantageous information of medium-wave infrared, LONG WAVE INFRARED has been aggravated.
Description
Technical field
The invention belongs to image computer technical fields, and in particular to a kind of multi-band image fusion identification method, the party
Method is mainly used for working in multi-spectral imaging detection system.
Background technology
The overwhelming majority of current so-called multi-band image fusion identification method is based on dual-band image.Common combination
Have：Visible ray and millimeter-wave image fusion, it is seen that light is merged with infrared image, the combinations such as infrared different-waveband image co-registration.This
Kind is so-called to be based on multiwave image co-registration recognition methods, it is impossible to be known as multi-band image fusion recognition side truly
Method.Three kinds and the identification of band above image co-registration are applied directly to based on two kinds of Band fusion image-recognizing methods, can be brought very
More problems.
The content of the invention
The technical problem to be solved in the present invention is to provide objective attribute target attribute fusion identification method in a kind of multi-spectral imaging system,
Blending algorithm proposed by the present invention is the fusion recognition algorithm of LONG WAVE INFRARED image, medium-wave infrared image and visible images, is
Truly multi-band image fusion recognition algorithm.The multi-band image fusion identification method of the present invention causes to image data
Processing, more interference-free to influence, imaging effect is more preferable.
The present invention is achieved through the following technical solutions：
A kind of multi-band image fusion identification method, comprises the following steps：
A1：Sample training is carried out respectively with the training sample of medium-wave infrared image, LONG WAVE INFRARED image, visible images
Afterwards, generation medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible images grader are corresponded respectively；
A2：Medium-wave infrared image, LONG WAVE INFRARED image are subjected to Fusion Features, then fusion figure is obtained after image reconstruction
Picture, and blending image is merged to form Fusion training sample with training sample；
A3：Fusion is instructed using medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible images grader
Practice sample and carry out classification processing, and obtain correct decision rate of each grader each to every one kind, mistake differentiation rate, refuse to differentiate
Rate；
A4：By medium-wave infrared images, LONG WAVE INFRARED images, obtains target video frame medium-wave infrared in a manner of visible image capturing
Image, target video frame length ripple infrared image, target video frame visible images, and correspondence is sent to medium-wave infrared image respectively
Grader, LONG WAVE INFRARED Image Classifier are classified in visible images grader and extract target signature information, are divided
Class with the correct decision rate analysis of corresponding grader as a result, obtained the probability assignments of each grader by corresponding classification results
Function；
A5：After the probability distribution function of arbitrary 2 graders is selected to carry out conflict judgement, obtained with the mode of synthesis or selection
Probability distribution function among obtaining；
A6：Conflict after judgement with intermediate probability distribution function and the probability distribution function of remaining grader, with synthesis
Or the mode of selection obtains final probability distribution function；
A7：When final probability distribution function is more than threshold value, recognition result is exported.
The detailed process of A1 is：
A11：The corresponding acquisition medium wave of medium-wave infrared video camera, LONG WAVE INFRARED video camera, visible light camera is respectively adopted
Infrared image, LONG WAVE INFRARED image, visible images,
A12：Respectively by medium-wave infrared image, LONG WAVE INFRARED image, visible images each after pretreatment, target is extracted
Characteristic information, detection, the tracking of target are carried out according to target signature, and are obtained target area and formed training sample, training sample
After carrying out sample training, generation medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible ray figure are corresponded respectively
As grader.
The detailed process of A2 is：
A21：By medium-wave infrared image, LONG WAVE INFRARED image after image registration, filtering and noise reduction pretreatment, extraction target is special
Reference ceases, and carries out Fusion Features, then blending image is obtained after image reconstruction；
A22：Blending image is merged to form Fusion training sample with training sample.
The detailed process of A3 is：Load medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible images point
Class device makes a gift to someone Fusion training sample respectively to medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible images
It is visible to obtain Fusion training sample medium wave classification results, Fusion training sample long wave classification results, Fusion training sample for grader
It is respective to obtain medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible images grader difference for light classification results
Correct decision rate, mistake differentiation rate to every one kind refuse differentiation rate, are considered as other all kinds of mistakes per a kind of correct decision rate
Differentiation rate.
The detailed process of A4 is：
A41：The corresponding acquisition target of medium-wave infrared video camera, LONG WAVE INFRARED video camera, visible light camera is respectively adopted
Video frame medium-wave infrared image, target video frame length ripple infrared image, target video frame visible images, will be in target video frame
Ripple infrared image, target video frame length ripple infrared image, that target video frame visible images are sent to medium wave correspondingly is red
Outer Image Classifier, LONG WAVE INFRARED Image Classifier are classified in visible images grader and extract target signature information,
Obtain target video frame medium wave classification results, target video frame length ripple classification results, target video frame visible ray classification results；
A42：According to target video frame medium wave classification results and medium-wave infrared Image Classifier to the correct decision of every one kind
Rate, mistake differentiation rate, find corresponding medium wave probability distribution function；According to target video frame length ripple classification results and medium-wave infrared figure
As correct decision rate of the grader to every one kind, mistake differentiation rate, corresponding long wave probability distribution function is found, according to target video
Correct decision rate, the mistake differentiation rate of frame visible ray classification results and visible images grader to every one kind, finding correspondence can
See light probability distribution function.
The detailed process of A5 is：
A51：Conflict analysis judgement is carried out to medium wave probability distribution function and long wave probability distribution function, is turned if conflict
A53, do not conflict, turn A52；
A52：Synthesis medium wave probability distribution function and long wave probability distribution function form intermediate probability distribution function, then turn
A54、A6；
A53：Compare the correct decision rate of medium-wave infrared Image Classifier and LONG WAVE INFRARED Image Classifier, if medium-wave infrared
The correct decision rate of Image Classifier is height, then medium wave probability distribution function is selected as intermediate probability distribution function, if long wave
The correct decision rate of infrared image grader is height, then long wave probability distribution function is as intermediate probability distribution function, then turns
A54、A6；
A54：The correct decision rate of medium-wave infrared Image Classifier and LONG WAVE INFRARED Image Classifier is weighted, is put down
, as intermediate correct decision rate after summing, A6 is turned；
The detailed process of A6 is：
A61：Conflict analysis judgement is carried out to intermediate probability distribution function and visible ray probability distribution function, is turned if conflict
A63, do not conflict, turn A62；
A62：Probability distribution function and visible ray probability distribution function form final probability distribution function among synthesis, then turn
A7；
A63：Compare the correct decision rate of visible images grader and intermediate correct decision rate, if intermediate correct decision rate
For height, then intermediate probability distribution function is selected as final probability distribution function, if the correct decision of visible images grader
Rate is height, then, it is seen that the probability distribution function corresponding to light image grader is as final probability distribution function, then turns A7.
The detailed process of A7 is：When final probability distribution function is more than threshold value, then in target video frame medium-wave infrared figure
Target signature information is superimposed on picture, target video frame length ripple infrared image, target video frame visible images and obtains final figure
Picture, final image and objective attribute target attribute classification results are sent to display output module, are shown and exported recognition result, mesh respectively
Mark characteristic information includes：Target sizes, profile, texture, boundary rectangle information.
Synthesis medium wave probability distribution function and long wave probability distribution function form the detailed process of intermediate probability distribution function
For：Medium wave probability distribution function and long wave probability distribution function are obtained, according to the probability point for correcting D-S, Bayes, fuzzy reasoning
With function composition rule, intermediate probability distribution function is obtained.
Probability distribution function and visible ray probability distribution function form the specific mistake of final probability distribution function among synthesis
Cheng Wei：Obtain intermediate probability distribution function and visible ray probability distribution function, according to correct D-S, Bayes, fuzzy reasoning it is general
Rate partition function composition rule, obtains final probability distribution function.
The present invention design principle be：The sensor of different-waveband has different imaging characteristics, and the spectral coverage of imaging covers
The wave bands such as visible ray, millimeter wave, infrared light.Visible images have abundant details and color information and higher resolution
Rate, but easily influenced by conditions such as weather and times.The image resolution ratio of infrared band is low, details deficiency, but can be with whole day
Wait work, strong antijamming capability.For target, the contour feature of target is clearer in LONG WAVE INFRARED image, medium-wave infrared
The objective contour feature of image is unclear, but target upper temperatures area stereovision is stronger.It is red using long wave in the system of the present invention
Outside, medium-wave infrared and visible light sensor carry out fusion recognition, can make full use of the message complementary sense between image, pass through structure
To target object description more fully model, the level of understanding of enhancing image and the reliability of information are not only able to, additionally it is possible to improve
The detection probability of target and Target attribute recognition accuracy rate.And how to be based on LONG WAVE INFRARED, medium-wave infrared and visible light sensing
It is for the most key core, present invention research discovery that device, which carries out building rational fusion recognition model,：By medium-wave infrared image
It is merged with LONG WAVE INFRARED image, for supplementary training sample, is carried out on the basis of the training sample after supplement at data
Reason, therefore on the visible images with abundant details and color information and higher resolution ratio and can utilize
Medium-wave infrared image and the respective advantage of LONG WAVE INFRARED image carry out classifying and dividing, by the data fusion of respective advantage maximum, because
This, can maximumlly reduce the loss of information.The present invention builds 3 kinds of graders before this, and medium-wave infrared image and long wave are red
The blending image of outer image is as supplementary training sample, since blending image is that the feature utilized shifts to an earlier date integration technology so that former
The single training sample that begins is provided with the outstanding feature of the respective advantage of medium-wave infrared image and LONG WAVE INFRARED image, has aggravated medium wave
The probability of the respective advantageous information of infrared, LONG WAVE INFRARED, then utilizes the synthesis of probability function or the apparent probability letter of selective advantage
Number is used as final probability function, while when judging 3 kinds of probability functions is caused to be involved in information operation, is equivalent to higher
On the basis of image in different resolution, especially emphasize to highlight the design of infrared imaging advantage proportion.
Compared with prior art, the present invention it has the following advantages and advantages：A kind of multiband of the present invention
Image co-registration recognition methods, make full use of each band image image information and each band class device preliminary classification conclusion provide
Information, and the Basic Probability As-signment to various conclusions is translated into, the loss of information is reduced, emerging system assigns probability
Value is more rationalized, and solves composition problem during evidences conflict, improves target identification accuracy rate, meets multiband detection system
The requirement objective accuracy of identification demand of system.
Description of the drawings
Attached drawing described herein is used for providing further understanding the embodiment of the present invention, forms one of the application
Point, do not form the restriction to the embodiment of the present invention.In the accompanying drawings：
Fig. 1 is the flow chart of the present invention.
Specific embodiment
Understand to make the object, technical solutions and advantages of the present invention clearer, the present invention is made with reference to embodiment
Further to be described in detail, exemplary embodiment of the invention and its explanation are only used for explaining the present invention, are not intended as to this
The restriction of invention.
Embodiment one
As shown in Figure 1,
A kind of multi-band image fusion identification method, comprises the following steps：
A1：Sample training is carried out respectively with the training sample of medium-wave infrared image, LONG WAVE INFRARED image, visible images
Afterwards, generation medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible images grader are corresponded respectively；
A2：Medium-wave infrared image, LONG WAVE INFRARED image are subjected to Fusion Features, then fusion figure is obtained after image reconstruction
Picture, and blending image is merged to form Fusion training sample with training sample；
A3：Fusion is instructed using medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible images grader
Practice sample and carry out classification processing, and obtain correct decision rate of each grader each to every one kind, mistake differentiation rate, refuse to differentiate
Rate；
A4：By medium-wave infrared images, LONG WAVE INFRARED images, obtains target video frame medium-wave infrared in a manner of visible image capturing
Image, target video frame length ripple infrared image, target video frame visible images, and correspondence is sent to medium-wave infrared image respectively
Grader, LONG WAVE INFRARED Image Classifier are classified in visible images grader and extract target signature information, are divided
Class with the correct decision rate analysis of corresponding grader as a result, obtained the probability assignments of each grader by corresponding classification results
Function；
A5：After the probability distribution function of arbitrary 2 graders is selected to carry out conflict judgement, obtained with the mode of synthesis or selection
Probability distribution function among obtaining；
A6：Conflict after judgement with intermediate probability distribution function and the probability distribution function of remaining grader, with synthesis
Or the mode of selection obtains final probability distribution function；
A7：When final probability distribution function is more than threshold value, recognition result is exported.
The detailed process of A1 is：
A11：The corresponding acquisition medium wave of medium-wave infrared video camera, LONG WAVE INFRARED video camera, visible light camera is respectively adopted
Infrared image, LONG WAVE INFRARED image, visible images,
A12：Respectively by medium-wave infrared image, LONG WAVE INFRARED image, visible images each after pretreatment, target is extracted
Characteristic information, detection, the tracking of target are carried out according to target signature, and are obtained target area and formed training sample, training sample
After carrying out sample training, generation medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible ray figure are corresponded respectively
As grader.
The detailed process of A2 is：
A21：By medium-wave infrared image, LONG WAVE INFRARED image after image registration, filtering and noise reduction pretreatment, extraction target is special
Reference ceases, and carries out Fusion Features, then blending image is obtained after image reconstruction；
A22：Blending image is merged to form Fusion training sample with training sample.
The detailed process of A3 is：Load medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible images point
Class device makes a gift to someone Fusion training sample respectively to medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible images
It is visible to obtain Fusion training sample medium wave classification results, Fusion training sample long wave classification results, Fusion training sample for grader
It is respective to obtain medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible images grader difference for light classification results
Correct decision rate, mistake differentiation rate to every one kind refuse differentiation rate, are considered as other all kinds of mistakes per a kind of correct decision rate
Differentiation rate.
The detailed process of A4 is：
A41：The corresponding acquisition target of medium-wave infrared video camera, LONG WAVE INFRARED video camera, visible light camera is respectively adopted
Video frame medium-wave infrared image, target video frame length ripple infrared image, target video frame visible images, will be in target video frame
Ripple infrared image, target video frame length ripple infrared image, that target video frame visible images are sent to medium wave correspondingly is red
Outer Image Classifier, LONG WAVE INFRARED Image Classifier are classified in visible images grader and extract target signature information,
Obtain target video frame medium wave classification results, target video frame length ripple classification results, target video frame visible ray classification results；
A42：According to target video frame medium wave classification results and medium-wave infrared Image Classifier to the correct decision of every one kind
Rate, mistake differentiation rate, find corresponding medium wave probability distribution function；According to target video frame length ripple classification results and medium-wave infrared figure
As correct decision rate of the grader to every one kind, mistake differentiation rate, corresponding long wave probability distribution function is found, according to target video
Correct decision rate, the mistake differentiation rate of frame visible ray classification results and visible images grader to every one kind, finding correspondence can
See light probability distribution function.
The detailed process of A5 is：
A51：Conflict analysis judgement is carried out to medium wave probability distribution function and long wave probability distribution function, is turned if conflict
A53, do not conflict, turn A52；
A52：Synthesis medium wave probability distribution function and long wave probability distribution function form intermediate probability distribution function, then turn
A54、A6；
A53：Compare the correct decision rate of medium-wave infrared Image Classifier and LONG WAVE INFRARED Image Classifier, if medium-wave infrared
The correct decision rate of Image Classifier is height, then medium wave probability distribution function is selected as intermediate probability distribution function, if long wave
The correct decision rate of infrared image grader is height, then long wave probability distribution function is as intermediate probability distribution function, then turns
A54、A6；
A54：The correct decision rate of medium-wave infrared Image Classifier and LONG WAVE INFRARED Image Classifier is weighted, is put down
, as intermediate correct decision rate after summing, A6 is turned；
The detailed process of A6 is：
A61：Conflict analysis judgement is carried out to intermediate probability distribution function and visible ray probability distribution function, is turned if conflict
A63, do not conflict, turn A62；
A62：Probability distribution function and visible ray probability distribution function form final probability distribution function among synthesis, then turn
A7；
A63：Compare the correct decision rate of visible images grader and intermediate correct decision rate, if intermediate correct decision rate
For height, then intermediate probability distribution function is selected as final probability distribution function, if the correct decision of visible images grader
Rate is height, then, it is seen that the probability distribution function corresponding to light image grader is as final probability distribution function, then turns A7.
The detailed process of A7 is：When final probability distribution function is more than threshold value, then in target video frame medium-wave infrared figure
Target signature information is superimposed on picture, target video frame length ripple infrared image, target video frame visible images and obtains final figure
Picture, final image and objective attribute target attribute classification results are sent to display output module, are shown and exported recognition result, mesh respectively
Mark characteristic information includes：Target sizes, profile, texture, boundary rectangle information.
Synthesis medium wave probability distribution function and long wave probability distribution function form the detailed process of intermediate probability distribution function
For：Medium wave probability distribution function and long wave probability distribution function are obtained, according to the probability point for correcting D-S, Bayes, fuzzy reasoning
With function composition rule, intermediate probability distribution function is obtained.
Probability distribution function and visible ray probability distribution function form the specific mistake of final probability distribution function among synthesis
Cheng Wei：Obtain intermediate probability distribution function and visible ray probability distribution function, according to correct D-S, Bayes, fuzzy reasoning it is general
Rate partition function composition rule, obtains final probability distribution function.
As shown in Figure 1,
When extracting characteristics of image, it is necessary to rational characteristic information, therefore image zooming-out is given in the present embodiment Fig. 1
The process of characteristic information is as follows：
Image to be identified is read in S1, initialization
S2, by target's feature-extraction in carry out image,
The respective calculation of correlation benchmark of S3 and chosen feature,
S4, the respective image identification evidence collection of generation,
S5, by calculate image target area entropy,
S6, target area entropy detection is carried out, when entropy is less than threshold value, then goes to step S2 and extract more target signatures
Information；Otherwise step S7 is performed,
S7, image target area gray consistency is calculated, and carries out target area gray consistency detection, work as local gray level
When uniformity is less than threshold value, then goes to step S2 and extract more target signature informations.
Above-described specific embodiment has carried out the purpose of the present invention, technical solution and advantageous effect further
It is described in detail, it should be understood that the foregoing is merely the specific embodiments of the present invention, is not intended to limit the present invention
Protection domain, within the spirit and principles of the invention, any modification, equivalent substitution, improvement and etc. done should all include
Within protection scope of the present invention.
Claims (9)
1. a kind of multi-band image fusion identification method, which is characterized in that comprise the following steps：
A1：After sample training being carried out with the training sample of medium-wave infrared image, LONG WAVE INFRARED image, visible images respectively, point
Medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible images grader Yi Yiduiying not generated；
A2：Medium-wave infrared image, LONG WAVE INFRARED image are carried out Fusion Features, then blending image obtained after image reconstruction, and
Blending image is merged to form Fusion training sample with training sample；
A3：Using medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible images grader to Fusion training sample
This carries out classification processing, and obtains correct decision rate of each grader each to every one kind, mistake differentiation rate, refuses differentiation rate；
A4：By medium-wave infrared images, LONG WAVE INFRARED images, obtains target video frame medium-wave infrared figure in a manner of visible image capturing
Picture, target video frame length ripple infrared image, target video frame visible images, and the corresponding medium-wave infrared image that is sent to divides respectively
Class device, LONG WAVE INFRARED Image Classifier are classified in visible images grader and extract target signature information, are classified
As a result, the probability assignments letter of each grader is obtained with the correct decision rate analysis of corresponding grader by corresponding classification results
Number；
A5：After the probability distribution function of arbitrary 2 graders is selected to carry out conflict judgement, in being obtained with the mode of synthesis or selection
Between probability distribution function；
A6：Conflict after judgement with intermediate probability distribution function and the probability distribution function of remaining grader, with synthesis or choosing
The mode selected obtains final probability distribution function；
A7：When final probability distribution function is more than threshold value, recognition result is exported.
2. a kind of multi-band image fusion identification method according to claim 1, it is characterised in that：
The detailed process of A1 is：
A11：The corresponding acquisition medium-wave infrared of medium-wave infrared video camera, LONG WAVE INFRARED video camera, visible light camera is respectively adopted
Image, LONG WAVE INFRARED image, visible images,
A12：Respectively by medium-wave infrared image, LONG WAVE INFRARED image, visible images each after pretreatment, target signature is extracted
Information, detection, the tracking of target are carried out according to target signature, and are obtained target area and formed training sample, and training sample carries out
After sample training, generation medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible images point are corresponded respectively
Class device.
3. a kind of multi-band image fusion identification method according to claim 1, it is characterised in that：
The detailed process of A2 is：
A21：By medium-wave infrared image, LONG WAVE INFRARED image after image registration, filtering and noise reduction pretreatment, extraction target signature letter
Breath carries out Fusion Features, then blending image is obtained after image reconstruction；
A22：Blending image is merged to form Fusion training sample with training sample.
4. a kind of multi-band image fusion identification method according to claim 1, it is characterised in that：
The detailed process of A3 is：Load medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible images classification
Device makes a gift to someone Fusion training sample respectively to medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible images point
Class device obtains Fusion training sample medium wave classification results, Fusion training sample long wave classification results, Fusion training sample visible ray
It is each right respectively to obtain medium-wave infrared Image Classifier, LONG WAVE INFRARED Image Classifier, visible images grader for classification results
Every a kind of correct decision rate, mistake differentiation rate refuse differentiation rate, and being considered as other all kinds of mistakes per a kind of correct decision rate sentences
Not rate.
5. a kind of multi-band image fusion identification method according to any one in claim 1-4, it is characterised in that：
The detailed process of A4 is：
A41：The corresponding acquisition target video of medium-wave infrared video camera, LONG WAVE INFRARED video camera, visible light camera is respectively adopted
Frame medium-wave infrared image, target video frame length ripple infrared image, target video frame visible images, target video frame medium wave is red
Outer image, target video frame length ripple infrared image, target video frame visible images are sent to medium-wave infrared figure correspondingly
As being classified in grader, LONG WAVE INFRARED Image Classifier, visible images grader and extracting target signature information, obtain
Target video frame medium wave classification results, target video frame length ripple classification results, target video frame visible ray classification results；
A42：According to target video frame medium wave classification results and medium-wave infrared Image Classifier to the correct decision rate, mistake of every one kind
Erroneous judgement not rate, finds corresponding medium wave probability distribution function；According to target video frame length ripple classification results and medium-wave infrared image point
Correct decision rate, mistake differentiation rate of the class device to every one kind, find corresponding long wave probability distribution function, can according to target video frame
See correct decision rate, the mistake differentiation rate of light classification results and visible images grader to every one kind, find corresponding visible ray
Probability distribution function.
6. a kind of multi-band image fusion identification method according to claim 5, it is characterised in that：
The detailed process of A5 is：
A51：Conflict analysis judgement is carried out to medium wave probability distribution function and long wave probability distribution function, turns A53, no if conflict
Conflict then turns A52；
A52：Synthesis medium wave probability distribution function and long wave probability distribution function form intermediate probability distribution function, then turn A54,
A6；
A53：Compare the correct decision rate of medium-wave infrared Image Classifier and LONG WAVE INFRARED Image Classifier, if medium-wave infrared image
The correct decision rate of grader is height, then medium wave probability distribution function is selected as intermediate probability distribution function, if LONG WAVE INFRARED
The correct decision rate of Image Classifier is height, then long wave probability distribution function is as intermediate probability distribution function, then turns A54, A6；
A54：The correct decision rate of medium-wave infrared Image Classifier and LONG WAVE INFRARED Image Classifier is weighted, be averaged, is asked
Intermediate correct decision rate is used as with rear, turns A6；
The detailed process of A6 is：
A61：Conflict analysis judgement is carried out to intermediate probability distribution function and visible ray probability distribution function, turn if conflict A63,
Do not conflict, turn A62；
A62：Probability distribution function and visible ray probability distribution function form final probability distribution function among synthesis, then turn A7；
A63：Compare the correct decision rate of visible images grader and intermediate correct decision rate, if intermediate correct decision rate is
Height then selects intermediate probability distribution function as final probability distribution function, if the correct decision rate of visible images grader
For height, then, it is seen that the probability distribution function corresponding to light image grader is as final probability distribution function, then turns A7.
7. a kind of multi-band image fusion identification method according to any one in claim 1-4, it is characterised in that：
The detailed process of A7 is：When final probability distribution function is more than threshold value, then in target video frame medium-wave infrared image, mesh
It is superimposed target signature information on mark video frame LONG WAVE INFRARED image, target video frame visible images and obtains final image, most
Whole image and objective attribute target attribute classification results are sent to display output module, are shown and exported recognition result respectively, and target is special
Reference breath includes：Target sizes, profile, texture, boundary rectangle information.
8. a kind of multi-band image fusion identification method according to claim 6, it is characterised in that：Synthesize medium wave probability point
The detailed process that intermediate probability distribution function is formed with function and long wave probability distribution function is：Obtain medium wave probability distribution function
With long wave probability distribution function, according to the probability distribution function composition rule for correcting D-S, Bayes, fuzzy reasoning, centre is obtained
Probability distribution function.
9. a kind of multi-band image fusion identification method according to claim 6, it is characterised in that：Probability point among synthesis
The detailed process that final probability distribution function is formed with function and visible ray probability distribution function is：Obtain intermediate probability assignments letter
Number and visible ray probability distribution function according to the probability distribution function composition rule for correcting D-S, Bayes, fuzzy reasoning, obtain
Final probability distribution function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711332225.7A CN108052976B (en) | 2017-12-13 | 2017-12-13 | Multiband image fusion identification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711332225.7A CN108052976B (en) | 2017-12-13 | 2017-12-13 | Multiband image fusion identification method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108052976A true CN108052976A (en) | 2018-05-18 |
| CN108052976B CN108052976B (en) | 2021-04-06 |
Family
ID=62132639
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711332225.7A Active CN108052976B (en) | 2017-12-13 | 2017-12-13 | Multiband image fusion identification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108052976B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108921803A (en) * | 2018-06-29 | 2018-11-30 | 华中科技大学 | A kind of defogging method based on millimeter wave and visual image fusion |
| CN109492714A (en) * | 2018-12-29 | 2019-03-19 | 同方威视技术股份有限公司 | Image processing apparatus and its method |
| CN110987189A (en) * | 2019-11-21 | 2020-04-10 | 北京都是科技有限公司 | Method, system and device for detecting temperature of target object |
| CN111401321A (en) * | 2020-04-17 | 2020-07-10 | Oppo广东移动通信有限公司 | Object recognition model training method and device, electronic equipment and readable storage medium |
| CN112070111A (en) * | 2020-07-28 | 2020-12-11 | 浙江大学 | Multi-target detection method and system adaptive to multiband images |
| CN113762277A (en) * | 2021-09-09 | 2021-12-07 | 东北大学 | Multi-band infrared image fusion method based on Cascade-GAN |
| CN114359743A (en) * | 2022-03-21 | 2022-04-15 | 华中科技大学 | Low-slow small target identification method and system based on multiband |
| CN114417691A (en) * | 2021-11-01 | 2022-04-29 | 西安电子科技大学 | Infrared band expanding method based on generation countermeasure network |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102592134A (en) * | 2011-11-28 | 2012-07-18 | 北京航空航天大学 | Multistage decision fusing and classifying method for hyperspectrum and infrared data |
| CN103984936A (en) * | 2014-05-29 | 2014-08-13 | 中国航空无线电电子研究所 | Multi-sensor multi-feature fusion recognition method for three-dimensional dynamic target recognition |
-
2017
- 2017-12-13 CN CN201711332225.7A patent/CN108052976B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102592134A (en) * | 2011-11-28 | 2012-07-18 | 北京航空航天大学 | Multistage decision fusing and classifying method for hyperspectrum and infrared data |
| CN103984936A (en) * | 2014-05-29 | 2014-08-13 | 中国航空无线电电子研究所 | Multi-sensor multi-feature fusion recognition method for three-dimensional dynamic target recognition |
Non-Patent Citations (2)
| Title |
|---|
| HASAN DEMIREL ET AL.: "Pose Invariant Face Recognition Using Probability Distribution Functions in Different Color Channels", 《IEEE SIGNAL PROCESSING LETTERS》 * |
| 王凤朝 等: "基于模糊证据理论的多特征目标融合检测算法", 《光学学报》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108921803A (en) * | 2018-06-29 | 2018-11-30 | 华中科技大学 | A kind of defogging method based on millimeter wave and visual image fusion |
| CN109492714A (en) * | 2018-12-29 | 2019-03-19 | 同方威视技术股份有限公司 | Image processing apparatus and its method |
| CN109492714B (en) * | 2018-12-29 | 2023-09-15 | 同方威视技术股份有限公司 | Image processing apparatus and method thereof |
| CN110987189A (en) * | 2019-11-21 | 2020-04-10 | 北京都是科技有限公司 | Method, system and device for detecting temperature of target object |
| CN111401321A (en) * | 2020-04-17 | 2020-07-10 | Oppo广东移动通信有限公司 | Object recognition model training method and device, electronic equipment and readable storage medium |
| CN112070111A (en) * | 2020-07-28 | 2020-12-11 | 浙江大学 | Multi-target detection method and system adaptive to multiband images |
| CN112070111B (en) * | 2020-07-28 | 2023-11-28 | 浙江大学 | Multi-target detection method and system adapting to multi-band image |
| CN113762277A (en) * | 2021-09-09 | 2021-12-07 | 东北大学 | Multi-band infrared image fusion method based on Cascade-GAN |
| CN113762277B (en) * | 2021-09-09 | 2024-05-24 | 东北大学 | Multiband infrared image fusion method based on Cascade-GAN |
| CN114417691A (en) * | 2021-11-01 | 2022-04-29 | 西安电子科技大学 | Infrared band expanding method based on generation countermeasure network |
| CN114359743A (en) * | 2022-03-21 | 2022-04-15 | 华中科技大学 | Low-slow small target identification method and system based on multiband |
| CN114359743B (en) * | 2022-03-21 | 2022-06-21 | 华中科技大学 | Low-slow small target identification method and system based on multiband |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108052976B (en) | 2021-04-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108052976A (en) | A kind of multi-band image fusion identification method | |
| Frome et al. | Large-scale privacy protection in google street view | |
| US9104914B1 (en) | Object detection with false positive filtering | |
| WO2021036267A1 (en) | Image detection method and related device | |
| KR20160143494A (en) | Saliency information acquisition apparatus and saliency information acquisition method | |
| CN103927741A (en) | SAR image synthesis method for enhancing target characteristics | |
| Shih et al. | Automatic reference color selection for adaptive mathematical morphology and application in image segmentation | |
| KR101778605B1 (en) | Method And Apparatus For Recognizing Vehicle License Plate | |
| US9197860B2 (en) | Color detector for vehicle | |
| CN112115979B (en) | Fusion method and device of infrared image and visible image | |
| Tschentscher et al. | Video-based parking space detection | |
| CN106815587A (en) | Image processing method and device | |
| Kowkabi et al. | A fast spatial–spectral preprocessing module for hyperspectral endmember extraction | |
| CN108492288B (en) | Random forest based multi-scale layered sampling high-resolution satellite image change detection method | |
| CN109886195A (en) | Skin identification method based on depth camera near-infrared single color gradation figure | |
| CN110910497B (en) | Method and system for realizing augmented reality map | |
| Jwaid et al. | Study and analysis of copy-move & splicing image forgery detection techniques | |
| CN103680145B (en) | A kind of people's car automatic identifying method based on local image characteristics | |
| Wang et al. | Saliency detection using mutual consistency-guided spatial cues combination | |
| Parande et al. | Concealed weapon detection in a human body by infrared imaging | |
| CN107832793A (en) | The sorting technique and system of a kind of high spectrum image | |
| CN109543610B (en) | Vehicle detection tracking method, device, equipment and storage medium | |
| CN111881924A (en) | Dim light vehicle illumination identification method combining illumination invariance and short-exposure illumination enhancement | |
| CN116403123A (en) | Remote sensing image change detection method based on deep convolution network | |
| TWI589468B (en) | Pedestrian detecting system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20210618 Address after: 621000 building 31, No.7, Section 2, Xianren Road, Youxian District, Mianyang City, Sichuan Province Patentee after: China Ordnance Equipment Group Automation Research Institute Co.,Ltd. Address before: 621000 Mianyang province Sichuan City Youxian District Road No. 7 two immortals Patentee before: China Ordnance Equipment Group Automation Research Institute |