CN103559715A - Abnormal detection method and device for hyper-spectral image - Google Patents
Abnormal detection method and device for hyper-spectral image Download PDFInfo
- Publication number
- CN103559715A CN103559715A CN201310551785.7A CN201310551785A CN103559715A CN 103559715 A CN103559715 A CN 103559715A CN 201310551785 A CN201310551785 A CN 201310551785A CN 103559715 A CN103559715 A CN 103559715A
- Authority
- CN
- China
- Prior art keywords
- detected
- pels
- collection
- abnormality detection
- pixel
- 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
Images
Abstract
The invention provides an abnormal detection method and device for a hyper-spectral image. The method includes the steps that the hyper-spectral image to be detected is obtained, abnormal detection is performed on the hyper-spectral image to be detected through an RX algorithm, so that an abnormal detection result is obtained, threshold segmentation is performed on the abnormal detection result to obtain a target image element set and a background image element set, the mean vector and the covariance matrix of the target image element set are calculated, the mean vector and the covariance matrix of the background image element set are calculated, the distance between each image element to be detected in the hyper-spectral image and each background image element is calculated through the mean vector and the covariance matrix of the background image element set, the distance between each image element to be detected in the hyper-spectral image and each target image element is calculated through the mean vector and the covariance matrix of the target image element set, and the distance between each image element to be detected and each background image element is subtracted from the distance between each image element to be detected and each target image element to obtain the abnormal detection result. The method and device can improve the detection rate of abnormal objects and reduce the false alarm rate.
Description
Technical field
The present invention relates to abnormality detection technical field, relate in particular to a kind of method for detecting abnormality and device of high spectrum image.
Background technology
Abnormality detection is a hot issue in high spectrum field.Abnormal object refers to the interested atural object of people.It has two large features, and one, has obvious SPECTRAL DIVERSITY, its two, the probability occurring in image is lower.
In prior art, conventionally use RX algorithm to realize abnormality detection.RX algorithm hypothesis image background is obeyed multivariate normal distribution, by calculating mean vector and the covariance matrix of backdrop pels collection, describes background information.Inventor finds in realizing the process of the invention: RX algorithm is estimated background information accurately, causes testing result to have higher false alarm rate.
Summary of the invention
In view of this, the invention provides a kind of method for detecting abnormality and device of high spectrum image, in order to solve in prior art accurately estimated background information of RX algorithm, cause testing result to have the problem of higher false alarm rate, its technical scheme is as follows:
On the one hand, a kind of method for detecting abnormality of high spectrum image, comprising:
Obtain high spectrum image to be detected;
Adopt RX algorithm to carry out abnormality detection to described high spectrum image to be detected, obtain abnormality detection result;
Described abnormality detection result is carried out to Threshold segmentation, obtain goal pels collection and backdrop pels collection;
Calculate mean vector and the covariance matrix of described goal pels collection, and the mean vector of described backdrop pels collection and covariance matrix;
For each pixel to be detected in described high spectrum image, mean vector by described backdrop pels collection and covariance matrix calculate the distance of described pixel to be detected and backdrop pels, mean vector by described goal pels collection and covariance matrix calculate the distance of described pixel to be detected and goal pels, the distance that deducts described pixel to be detected and goal pels by the distance of described pixel to be detected and backdrop pels, obtains abnormality detection result.
Wherein, use RX algorithm to carry out abnormality detection to described high spectrum image to be detected, obtain abnormality detection result matrix, comprising:
Utilize mean vector and the covariance matrix of described high spectrum image background to be detected, by A (i, j)=[r (i, j)-μ]
tΣ
-1[r (i, j)-μ] calculate described abnormality detection result matrix, wherein, A (i, j) is described abnormality detection result matrix, r (i, j) be at position (i in described high spectrum image to be detected, j) pixel, μ is the mean vector of background in described high spectrum image to be detected, Σ is the covariance matrix of background in described high spectrum image to be detected.
Wherein, described abnormality detection result is carried out to Threshold segmentation, obtains goal pels collection and backdrop pels collection, comprising:
Obtain confidence factor γ
lowerand γ
upper, and the grey level histogram of described abnormality detection result;
Grey level histogram by described abnormality detection result is determined and described confidence factor γ
lowercorresponding gray threshold TH
lower, and with described confidence factor γ
uppercorresponding gray threshold TH
upper;
From interval [TH
lower, TH
upper] in determine optimum gradation threshold value TH
best;
Described optimum gradation threshold value TH will be greater than in described abnormality detection result
bestpixel be defined as goal pels, obtain described goal pels collection, the pixel the goal pels except determining is defined as to backdrop pels, obtain described backdrop pels collection.
Wherein, described from interval [TH
lower, TH
upper] in determine optimum gradation threshold value TH
best, comprising:
By described interval [TH
lower, TH
upper] be divided into a plurality of isometric intervals;
Determine the quantity that falls into each interval abnormality detection result value in described abnormality detection result;
By the quantity that falls into each interval abnormality detection result value of determining, by preset rules, determine described optimum gradation threshold value TH
best;
Wherein, described preset rules is:
t
ifor falling into the quantity of i interval abnormality detection result value, T
i+1for falling into the quantity of i+1 interval abnormality detection result value.
Wherein, the mean vector by described backdrop pels collection and covariance matrix calculate the distance that described pixel to be detected is backdrop pels, comprising:
Utilize mean vector and the covariance matrix of described backdrop pels collection, by
calculate the distance of described pixel to be detected and backdrop pels, wherein, μ
0for the mean vector of described backdrop pels collection, Σ
0for the covariance matrix of described backdrop pels collection, D
0(i, j) is the distance of described pixel r to be detected (i, j) and backdrop pels;
Mean vector by described goal pels collection and covariance matrix calculate the distance that described pixel to be detected is goal pels and comprise:
Utilize mean vector and the covariance matrix of described goal pels collection, by
calculate the distance of described pixel to be detected and goal pels, wherein, r (i, j) is pixel to be detected, μ
1for the mean vector of described goal pels collection, Σ
1for the covariance matrix of described goal pels collection, D
1(i, j) is the distance of described pixel r to be detected (i, j) and goal pels.
An abnormal detector for high spectrum image, comprising:
Image collection module, for obtaining high spectrum image to be detected;
First detection module, for adopting RX algorithm to carry out abnormality detection to described high spectrum image to be detected, obtains abnormality detection result;
Threshold segmentation module, for described abnormality detection result is carried out to Threshold segmentation, obtains goal pels collection and backdrop pels collection;
Computing module, for calculating mean vector and the covariance matrix of described goal pels collection, and the mean vector of described backdrop pels collection and covariance matrix;
The second detection module, for each pixel to be detected for described high spectrum image, mean vector by described backdrop pels collection and covariance matrix calculate the distance of described pixel to be detected and backdrop pels, mean vector by described goal pels collection and covariance matrix calculate the distance of described pixel to be detected and goal pels, the distance that deducts described pixel to be detected and goal pels by the distance of described pixel to be detected and backdrop pels, obtains abnormality detection result.
Wherein, described first detection module comprises:
The first calculating sub module, for utilizing mean vector and the covariance matrix of described high spectrum image background to be detected, by A (i, j)=[r (i, j)-μ]
tΣ
-1[r (i, j)-μ] calculate described abnormality detection result matrix, wherein, A (i, j) is described abnormality detection result matrix, r (i, j) be at position (i in described high spectrum image to be detected, j) pixel, μ is the mean vector of background in described high spectrum image to be detected, Σ is the covariance matrix of background in described high spectrum image to be detected.
Wherein, described Threshold segmentation module comprises:
Obtain submodule, for obtaining confidence factor γ
lowerand γ
upper, and, the grey level histogram of described abnormality detection result;
First determines submodule, for the grey level histogram by described abnormality detection result, determines and described confidence factor γ
lowercorresponding gray threshold TH
lower, and with described confidence factor γ
uppercorresponding gray threshold TH
upper;
Second determines submodule, for from interval [TH
lower, TH
upper] in determine optimum gradation threshold value TH
best;
The 3rd determines submodule, for described abnormality detection result being greater than to described optimum gradation threshold value TH
bestpixel be defined as goal pels, obtain described goal pels collection, the pixel the goal pels except determining is defined as to backdrop pels, obtain described backdrop pels collection.
Wherein, described second determines that submodule comprises:
Interval division unit, for by described interval [TH
lower, TH
upper] be divided into a plurality of isometric intervals;
The first determining unit, for determining that described abnormality detection result falls into the quantity of each interval abnormality detection result value;
The second determining unit, for by the quantity that falls into each interval abnormality detection result value of determining, determines described optimum gradation threshold value TH by preset rules
best;
Wherein, described preset rules is:
t
ifor falling into the quantity of i interval abnormality detection result value, T
i+1for falling into the quantity of i+1 interval abnormality detection result value.
Wherein, described the second detection module comprises:
The second calculating sub module, for utilizing mean vector and the covariance matrix of described backdrop pels collection, by
calculate described pixel to be detected and be the distance with backdrop pels, wherein, μ
0for the mean vector of described backdrop pels collection, Σ
0for the covariance matrix of described backdrop pels collection, D
0(i, j) is the distance of described pixel r to be detected (i, j) and backdrop pels;
The 3rd calculating sub module, for utilizing mean vector and the covariance matrix of described goal pels collection, by
calculate the distance of described pixel to be detected and goal pels, wherein, r (i, j) is pixel to be detected, μ
1for the mean vector of described goal pels collection, Σ
1for the covariance matrix of described goal pels collection, D
1(i, j) is the distance of described pixel r to be detected (i, j) and goal pels.
Technique scheme has following beneficial effect:
The method for detecting abnormality of high spectrum image provided by the invention and device, after getting high spectrum image to be detected, first adopt RX algorithm to carry out abnormality detection to high spectrum image to be detected, then abnormality detection result is carried out to Threshold segmentation, obtain goal pels collection and backdrop pels collection, finally for each pixel to be detected, by goal pels collection and backdrop pels collection, determine the distance of pixel to be detected and goal pels, and the distance of pixel to be detected and backdrop pels, utilize the distance of pixel to be detected and backdrop pels to deduct the distance of pixel to be detected and goal pels, obtain the abnormality detection result value of this pixel to be detected, thereby obtain the abnormality detection result of high spectrum image to be detected.The method for detecting abnormality of high spectrum image provided by the invention and device, utilize the background information and the target information that in high spectrum image to be detected, contain to instruct follow-up abnormality detection work, the method is estimated background information more accurately, thereby make background information meet the hypothesis of RX algorithm to background information, and then improve abnormal object recall rate, reduction false alarm rate.
Accompanying drawing explanation
In order to be illustrated more clearly in the embodiment of the present invention or technical scheme of the prior art, to the accompanying drawing of required use in embodiment or description of the Prior Art be briefly described below, apparently, accompanying drawing in the following describes is only embodiments of the invention, for those of ordinary skills, do not paying under the prerequisite of creative work, other accompanying drawing can also be provided according to the accompanying drawing providing.
The schematic flow sheet of the method for detecting abnormality of a kind of high spectrum image that Fig. 1 provides for the embodiment of the present invention;
The schematic flow sheet of the method for detecting abnormality of the another kind of high spectrum image that Fig. 2 provides for the embodiment of the present invention;
The structural representation of the abnormal detector of a kind of high spectrum image that Fig. 3 provides for the embodiment of the present invention;
The structural representation of the abnormal detector of the another kind of high spectrum image that Fig. 4 provides for the embodiment of the present invention.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is clearly and completely described, obviously, described embodiment is only the present invention's part embodiment, rather than whole embodiment.Embodiment based in the present invention, those of ordinary skills, not making the every other embodiment obtaining under creative work prerequisite, belong to the scope of protection of the invention.
Refer to Fig. 1, the schematic flow sheet of the method for detecting abnormality of a kind of high spectrum image providing for the embodiment of the present invention, the method can comprise:
Step S101: obtain high spectrum image to be detected.
Wherein, high spectrum image is a three-dimensional array, concrete, and for M is capable, N is listed as, the image of a L wave band, high-spectrum image set image information and spectral information are.High spectrum image is comprised of M * N pixel, and each pixel r (i, j) is a L dimensional vector, and the numerical value by high spectrum image at L the wave band of (i, j) position forms.
Step S102: adopt RX algorithm to carry out abnormality detection to high spectrum image to be detected, obtain abnormality detection result.
Step S103: abnormality detection result is carried out to Threshold segmentation, obtain goal pels collection and backdrop pels collection.
Carrying out image threshold segmentation is a kind of traditional the most frequently used image partition method, because of its realize simple, calculated amount is little, performance compared with stable become image cut apart in the most widely used cutting techniques of fundamental sum.It is specially adapted to the image that target and background occupies different grey-scale scope.
Because abnormality detection of the prior art is the blind detection process of a unknown object information and background information, RX algorithm lacks available information background information is made to accurate estimation, and inaccurate background estimating causes testing result to have higher false alarm rate.In view of this, embodiment of the present invention passing threshold is cut apart and is obtained background information and the target information containing in high spectrum image to be detected, the background information of obtaining and target information are for the abnormality detection of aid in later, by background information and target information, can make accurate estimation to background, thereby improve recall rate, the reduction false alarm rate of follow-up abnormality detection.
Step S104: calculate mean vector and the covariance matrix of goal pels collection, and, the mean vector of backdrop pels collection and covariance matrix.
Step S105: for each pixel to be detected in high spectrum image, the distance that mean vector and covariance matrix by backdrop pels collection calculates pixel to be detected and backdrop pels, the distance that mean vector and covariance matrix by goal pels collection calculates pixel to be detected and goal pels, the distance that deducts pixel to be detected and goal pels by the distance of pixel to be detected and backdrop pels, obtains abnormality detection result.
The method for detecting abnormality of the high spectrum image that the embodiment of the present invention provides, after getting high spectrum image to be detected, first adopt RX algorithm to carry out abnormality detection to high spectrum image to be detected, then abnormality detection result is carried out to Threshold segmentation, obtain goal pels collection and backdrop pels collection, finally for each pixel to be detected, by goal pels collection and backdrop pels collection, determine the distance of pixel to be detected and goal pels, and the distance of pixel to be detected and backdrop pels, by the distance of pixel to be detected and backdrop pels, deduct the distance of pixel to be detected and goal pels, obtain the abnormality detection result value of this pixel to be detected, thereby obtain the abnormality detection result of high spectrum image to be detected.The method for detecting abnormality of the high spectrum image that the embodiment of the present invention provides, utilize the background information and the target information that in high spectrum image to be detected, contain to instruct follow-up abnormality detection work, the method is estimated background information more accurately, thereby make background information meet the hypothesis of RX algorithm to background information, and then improve abnormal object recall rate, reduction false alarm rate.
Refer to Fig. 2, the schematic flow sheet of the method for detecting abnormality of the another kind of high spectrum image providing for the embodiment of the present invention, the method can comprise:
Step S201: obtain high spectrum image R to be detected.
Wherein, high spectrum image R is a three-dimensional array, concrete, and for M is capable, N is listed as, the image of a L wave band, high-spectrum image set image information and spectral information are.High spectrum image is comprised of M * N pixel, and each pixel r (i, j) is a L dimensional vector, and the numerical value by high spectrum image at L the wave band of (i, j) position forms.
Step S202: the mean vector μ that calculates background in high spectrum image R to be detected.
Wherein, in high spectrum image R to be detected, the mean vector μ of background can calculate by through type (1):
Step S203: utilize the mean vector μ of background to calculate the covariance matrix Σ of background in high spectrum image R to be detected.
Wherein, in high spectrum image R to be detected, the covariance matrix Σ of background can calculate by through type (2):
Step S204: utilize mean vector μ and the covariance matrix Σ of background in high spectrum image R to be detected, adopt RX algorithm to carry out abnormality detection to high spectrum image R to be detected, obtain abnormality detection result matrix A.
Wherein, abnormality detection result matrix A is that M is capable, N is listed as, the gray level image of 1 wave band.
Adopting RX algorithm to carry out abnormality detection to high spectrum image R to be detected is specially: can calculate testing result matrix A by through type (3):
A(i,j)=[r(i,j)-μ]
TΣ
-1[r(i,j)-μ] (3)
Step S205: obtain confidence factor γ
lowerand γ
upper, and the grey level histogram of abnormality detection result, by the grey level histogram of abnormality detection result, determine and confidence factor γ
lowercorresponding gray threshold TH
lower, and with confidence factor γ
uppercorresponding gray threshold TH
upper.
Wherein, the probability that in confidence factor presentation video, background occurs.Concrete, confidence factor γ
lowerthe minimum probability that in presentation video, background occurs, confidence factor γ
upperthe maximum probability that in presentation video, background occurs.
Step S206: from interval [TH
lower, TH
upper] in determine optimum gradation threshold value TH
best.
In a kind of possible implementation, from interval [TH
lower, TH
upper] in determine optimum gradation threshold value TH
bestcan comprise: by interval [TH
lower, TH
upper] be divided into a plurality of isometric intervals; Determine the quantity that falls into each interval abnormality detection result value in abnormality detection result; By the quantity that falls into each interval abnormality detection result value of determining, by preset rules, determine optimum gradation threshold value TH
best.Wherein, preset rules can be:
t
ifor falling into the quantity of i interval abnormality detection result value, T
i+1for falling into the quantity of i+1 interval abnormality detection result value, i and
for being more than or equal to 1 integer.
Step S207: will be greater than optimum gradation threshold value TH in abnormality detection result
bestpixel be defined as goal pels, obtain goal pels collection H
1, the pixel the goal pels except determining is defined as to backdrop pels, obtain backdrop pels collection H
0.
Carrying out image threshold segmentation is a kind of traditional the most frequently used image partition method, because of its realize simple, calculated amount is little, performance compared with stable become image cut apart in the most widely used cutting techniques of fundamental sum.It is specially adapted to the image that target and background occupies different grey-scale scope.
Because abnormality detection of the prior art is the blind detection process of a unknown object information and background information, RX algorithm lacks available information background information is made to accurate estimation, and inaccurate background estimating causes testing result to have higher false alarm rate.In view of this, embodiment of the present invention passing threshold is cut apart and is obtained background information and the target information containing in high spectrum image to be detected, and the background information of obtaining and target information are for the abnormality detection of aid in later.On the one hand, background extraction information and target information are as priori, with the pixel estimated background more similar to background, can suppress abnormal object and noise, make background better meet algorithm hypothesis, on the other hand, the introducing of background information and target information can highlight target, thereby improves abnormal object recall rate, reduction false alarm rate.
Step S208: calculate goal pels collection H
1mean vector μ
1with covariance matrix Σ
1, and, backdrop pels collection H
0mean vector μ
0with covariance matrix Σ
0.
Step S209: for each pixel to be detected in high spectrum image R to be detected, by backdrop pels collection H
0mean vector μ
0with covariance matrix Σ
0calculate the distance of pixel to be detected and backdrop pels, by goal pels collection H
1mean vector μ
1with covariance matrix Σ
1calculate the distance of pixel to be detected and goal pels, by the distance of pixel to be detected and backdrop pels, deduct the distance of pixel to be detected and goal pels, obtain abnormality detection result.
For pixel r to be detected (i, j), by backdrop pels collection H
0mean vector μ
0with covariance matrix Σ
0calculating pixel r to be detected (i, j) is specially with the distance of backdrop pels: utilize backdrop pels collection H
0mean vector μ
0with covariance matrix Σ
0through type (4) calculates the distance D of pixel r to be detected (i, j) and backdrop pels
0(i, j):
Same, for pixel r to be detected (i, j), by goal pels collection H
1mean vector μ
1with covariance matrix Σ
1calculating pixel r to be detected (i, j) is specially with the distance of goal pels: utilize goal pels collection H
1mean vector μ
1with covariance matrix Σ
1, through type (5) calculates the distance D of pixel r to be detected (i, j) and goal pels
1(i, j):
Calculating the distance P of pixel r to be detected (i, j) with backdrop pels
0(i, j), and the distance P of pixel r to be detected (i, j) and goal pels
1after (i, j), through type (6) calculates the abnormality detection result value Δ D of pixel r (i, j)
0-1(i, j):
ΔD
0-1(i,j)=D
0(i,j)-D
1(i,j) (6)
For each pixel in high spectrum image to be detected, carry out aforesaid operations, just obtain the abnormality detection result Δ D of high spectrum image to be detected
0-1.
The method for detecting abnormality of the high spectrum image that the embodiment of the present invention provides, after getting high spectrum image to be detected, first adopt RX algorithm to carry out abnormality detection to high spectrum image to be detected, then abnormality detection result is carried out to Threshold segmentation, obtain goal pels collection and backdrop pels collection, finally for each pixel to be detected, by goal pels collection and backdrop pels collection, determine the distance of pixel to be detected and goal pels, and the distance of pixel to be detected and backdrop pels, utilize the distance of pixel to be detected and backdrop pels to deduct the distance of pixel to be detected and goal pels, obtain the abnormality detection result value of this pixel to be detected, thereby obtain the abnormality detection result of high spectrum image to be detected.The method for detecting abnormality of the high spectrum image that the embodiment of the present invention provides, utilize the background information and the target information that in high spectrum image to be detected, contain to instruct follow-up abnormality detection work, the method is estimated background information more accurately, thereby make background information meet the hypothesis of RX algorithm to background information, and then improve abnormal object recall rate, reduction false alarm rate.
Refer to Fig. 3, the structural representation of the abnormal detector of a kind of high spectrum image providing for the embodiment of the present invention, this device can comprise: image collection module 301, first detection module 302, Threshold segmentation module 303, computing module 304, the second detection module 305.Wherein:
The second detection module 305, for each pixel to be detected for high spectrum image to be detected, the distance that mean vector and covariance matrix by backdrop pels collection calculates pixel to be detected and backdrop pels, the distance that mean vector and covariance matrix by goal pels collection calculates pixel to be detected and goal pels, the distance that deducts pixel to be detected and goal pels by the distance of pixel to be detected and backdrop pels, obtains abnormality detection result.
The abnormal detector of the high spectrum image that the embodiment of the present invention provides, after getting high spectrum image to be detected, first adopt RX algorithm to carry out abnormality detection to high spectrum image to be detected, then abnormality detection result is carried out to Threshold segmentation, obtain goal pels collection and backdrop pels collection, finally for each pixel to be detected, by goal pels collection and backdrop pels collection, determine the distance of pixel to be detected and goal pels, and the distance of pixel to be detected and backdrop pels, utilize the distance of pixel to be detected and backdrop pels to deduct the distance of pixel to be detected and goal pels, obtain the abnormality detection result value of this pixel to be detected, thereby obtain the abnormality detection result of high spectrum image to be detected.The abnormal detector of the high spectrum image that the embodiment of the present invention provides, utilize the background information and the target information that in high spectrum image to be detected, contain to instruct follow-up abnormality detection work, estimated background information more accurately, thereby make background information meet the hypothesis of RX algorithm to background information, and then improve abnormal object recall rate, reduction false alarm rate.
Refer to Fig. 4, the structural representation of the abnormal detector of the another kind of high spectrum image providing for the embodiment of the present invention, this device can comprise: image collection module 401, first detection module 402, Threshold segmentation module 403, computing module 404, the second detection module 405.Wherein:
Further, first detection module comprises 402: the first calculating sub module 4021.
The first calculating sub module, for utilizing mean vector and the covariance matrix of high spectrum image background to be detected, by A (i, j)=[r (i, j)-μ]
tΣ
-1[r (i, j)-μ] calculate abnormality detection result matrix, wherein, A (i, j) is abnormality detection result matrix, r (i, j) be at position (i in high spectrum image to be detected, j) pixel, μ is the mean vector of background in high spectrum image to be detected, Σ is the covariance matrix of background in high spectrum image to be detected.
Further, Threshold segmentation module 403 can comprise: obtain submodule 4031, first and determine submodule 4032, second definite submodule 4033 and the 3rd definite submodule 4034.Wherein:
Obtain submodule 4031, for obtaining confidence factor γ
lowerand γ
upper, and the grey level histogram of abnormality detection result.
First determines submodule 4032, for the grey level histogram by abnormality detection result, determines and confidence factor γ
lowercorresponding gray threshold TH
lower, and with confidence factor γ
uppercorresponding gray threshold TH
upper.
Second determines submodule 4033, for from interval [TH
lower, TH
upper] in determine optimum gradation threshold value TH
best.
The 3rd determines submodule 4034, for abnormality detection result being greater than to optimum gradation threshold value TH
bestpixel be defined as goal pels, obtain goal pels collection, the pixel the goal pels except determining is defined as to backdrop pels, obtain backdrop pels collection.
Further, second determines that submodule 4033 can comprise: interval division unit, the first determining unit and the second determining unit.Wherein, interval division unit, for by described interval [TH
lower, TH
upper] be divided into a plurality of isometric intervals; The first determining unit, for determining that described abnormality detection result falls into the quantity of each interval abnormality detection result value; The second determining unit, for by the quantity that falls into each interval abnormality detection result value of determining, determines described optimum gradation threshold value TH by preset rules
best, preset rules can be:
t
ifor falling into the quantity of i interval abnormality detection result value, T
i+1for falling into the quantity of i+1 interval abnormality detection result value.
The second detection module 405, for each pixel to be detected for high spectrum image to be detected, the distance that mean vector and covariance matrix by backdrop pels collection calculates pixel to be detected and backdrop pels, the distance that mean vector and covariance matrix by goal pels collection calculates pixel to be detected and goal pels, the distance that deducts pixel to be detected and goal pels by the distance of pixel to be detected and backdrop pels, obtains abnormality detection result.
Further, the second detection module 405 comprises: the second calculating sub module 4051 and the 3rd calculating sub module 4052.Wherein:
The second calculating sub module 4051, for utilizing mean vector and the covariance matrix of backdrop pels collection, by
calculate the distance of pixel to be detected and backdrop pels, wherein, μ
0for the mean vector of backdrop pels collection, Σ
0for the covariance matrix of backdrop pels collection, P
0(i, j) is the distance of pixel r to be detected (i, j) and backdrop pels.
The 3rd calculating sub module 4052, for utilizing mean vector and the covariance matrix of backdrop pels collection, by
calculate the distance of pixel to be detected and goal pels, wherein, r (i, j) is pixel to be detected, μ
1for the mean vector of goal pels collection, Σ
1for the covariance matrix of goal pels collection, D
1(i, j) is the distance of pixel r to be detected (i, j) and goal pels.
The abnormal detector of the high spectrum image that the embodiment of the present invention provides, after getting high spectrum image to be detected, first adopt RX algorithm to carry out abnormality detection to high spectrum image to be detected, then abnormality detection result is carried out to Threshold segmentation, obtain goal pels collection and backdrop pels collection, finally for each pixel to be detected, by goal pels collection and backdrop pels collection, determine the distance of pixel to be detected and goal pels, and the distance of pixel to be detected and backdrop pels, utilize the distance of pixel to be detected and backdrop pels to deduct the distance of pixel to be detected and goal pels, obtain the abnormality detection result value of this pixel to be detected, thereby obtain the abnormality detection result of high spectrum image to be detected.The abnormal detector of the high spectrum image that the embodiment of the present invention provides, utilize the background information and the target information that in high spectrum image to be detected, contain to instruct follow-up abnormality detection work, estimated background information more accurately, thereby make background information meet the hypothesis of RX algorithm to background information, and then improve abnormal object recall rate, reduction false alarm rate.
It should be noted that, each embodiment in this instructions all adopts the mode of going forward one by one to describe, and each embodiment stresses is the difference with other embodiment, between each embodiment identical similar part mutually referring to.For device or system class embodiment, because it is substantially similar to embodiment of the method, so description is fairly simple, relevant part is referring to the part explanation of embodiment of the method.
Also it should be noted that, in this article, relational terms such as the first and second grades is only used for an entity or operation to separate with another entity or operational zone, and not necessarily requires or imply and between these entities or operation, have the relation of any this reality or sequentially.And, term " comprises ", " comprising " or its any other variant are intended to contain comprising of nonexcludability, thereby the process, method, article or the equipment that make to comprise a series of key elements not only comprise those key elements, but also comprise other key elements of clearly not listing, or be also included as the intrinsic key element of this process, method, article or equipment.The in the situation that of more restrictions not, the key element being limited by statement " comprising ... ", and be not precluded within process, method, article or the equipment that comprises described key element and also have other identical element.
The software module that the method for describing in conjunction with embodiment disclosed herein or the step of algorithm can directly use hardware, processor to carry out, or the combination of the two is implemented.Software module can be placed in the storage medium of any other form known in random access memory (RAM), internal memory, ROM (read-only memory) (ROM), electrically programmable ROM, electrically erasable ROM, register, hard disk, moveable magnetic disc, CD-ROM or technical field.
Above-mentioned explanation to the disclosed embodiments, makes professional and technical personnel in the field can realize or use the present invention.To the multiple modification of these embodiment, will be apparent for those skilled in the art, General Principle as defined herein can, in the situation that not departing from the spirit or scope of the present invention, realize in other embodiments.Therefore, the present invention will can not be restricted to these embodiment shown in this article, but will meet the widest scope consistent with principle disclosed herein and features of novelty.
Claims (10)
1. a method for detecting abnormality for high spectrum image, is characterized in that, comprising:
Obtain high spectrum image to be detected;
Adopt RX algorithm to carry out abnormality detection to described high spectrum image to be detected, obtain abnormality detection result;
Described abnormality detection result is carried out to Threshold segmentation, obtain goal pels collection and backdrop pels collection;
Calculate mean vector and the covariance matrix of described goal pels collection, and the mean vector of described backdrop pels collection and covariance matrix;
For each pixel to be detected in described high spectrum image, mean vector by described backdrop pels collection and covariance matrix calculate the distance of described pixel to be detected and backdrop pels, mean vector by described goal pels collection and covariance matrix calculate the distance of described pixel to be detected and goal pels, the distance that deducts described pixel to be detected and goal pels by the distance of described pixel to be detected and backdrop pels, obtains abnormality detection result.
2. method according to claim 1, is characterized in that, uses RX algorithm to carry out abnormality detection to described high spectrum image to be detected, obtains abnormality detection result matrix, comprising:
Utilize mean vector and the covariance matrix of described high spectrum image background to be detected, by A (i, j)=[r (i, j)-μ]
tΣ
-1[r (i, j)-μ] calculate described abnormality detection result matrix, wherein, A (i, j) is described abnormality detection result matrix, r (i, j) be at position (i in described high spectrum image to be detected, j) pixel, μ is the mean vector of background in described high spectrum image to be detected, Σ is the covariance matrix of background in described high spectrum image to be detected.
3. method according to claim 1, is characterized in that, described abnormality detection result is carried out to Threshold segmentation, obtains goal pels collection and backdrop pels collection, comprising:
Obtain confidence factor γ
lowerand γ
upper, and the grey level histogram of described abnormality detection result;
Grey level histogram by described abnormality detection result is determined and described confidence factor γ
lowercorresponding gray threshold TH
lower, and with described confidence factor γ
uppercorresponding gray threshold TH
upper;
From interval [TH
lower, TH
upper] in determine optimum gradation threshold value TH
best;
Described optimum gradation threshold value TH will be greater than in described abnormality detection result
bestpixel be defined as goal pels, obtain described goal pels collection, the pixel the goal pels except determining is defined as to backdrop pels, obtain described backdrop pels collection.
4. method according to claim 3, is characterized in that, described from interval [TH
lower, TH
upper] in determine optimum gradation threshold value TH
best, comprising:
By described interval [TH
lower, TH
upper] be divided into a plurality of isometric intervals;
Determine the quantity that falls into each interval abnormality detection result value in described abnormality detection result;
By the quantity that falls into each interval abnormality detection result value of determining, by preset rules, determine described optimum gradation threshold value TH
best;
Wherein, described preset rules is:
t
ifor falling into the quantity of i interval abnormality detection result value, T
i+1for falling into the quantity of i+1 interval abnormality detection result value.
5. according to the method described in any one in claim 1-4, it is characterized in that, the mean vector by described backdrop pels collection and covariance matrix calculate the distance of described pixel to be detected and backdrop pels, comprising:
Utilize mean vector and the covariance matrix of described backdrop pels collection, by
calculate the distance of described pixel to be detected and backdrop pels, wherein, μ
0for the mean vector of described backdrop pels collection, Σ
0for the covariance matrix of described backdrop pels collection, D
0(i, j) is the distance of described pixel r to be detected (i, j) and backdrop pels;
By the mean vector of described goal pels collection and the distance of the covariance matrix described pixel to be detected of calculating and goal pels, comprise:
Utilize mean vector and the covariance matrix of described goal pels collection, by
calculate the distance of described pixel to be detected and goal pels, wherein, r (i, j) is pixel to be detected, μ
1for the mean vector of described goal pels collection, Σ
1for the covariance matrix of described goal pels collection, D
1(i, j) is the distance of described pixel r to be detected (i, j) and goal pels.
6. an abnormal detector for high spectrum image, is characterized in that, comprising:
Image collection module, for obtaining high spectrum image to be detected;
First detection module, for adopting RX algorithm to carry out abnormality detection to described high spectrum image to be detected, obtains abnormality detection result;
Threshold segmentation module, for described abnormality detection result is carried out to Threshold segmentation, obtains goal pels collection and backdrop pels collection;
Computing module, for calculating mean vector and the covariance matrix of described goal pels collection, and the mean vector of described backdrop pels collection and covariance matrix;
The second detection module, for each pixel to be detected for described high spectrum image, mean vector by described backdrop pels collection and covariance matrix calculate the distance of described pixel to be detected and backdrop pels, mean vector by described goal pels collection and covariance matrix calculate the distance of described pixel to be detected and goal pels, the distance that deducts described pixel to be detected and goal pels by the distance of described pixel to be detected and backdrop pels, obtains abnormality detection result.
7. device according to claim 6, is characterized in that, described first detection module comprises:
The first calculating sub module, for utilizing mean vector and the covariance matrix of described high spectrum image background to be detected, by A (i, j)=[r (i, j)-μ]
tΣ
-1[r (i, j)-μ] calculate described abnormality detection result matrix, wherein, A (i, j) is described abnormality detection result matrix, r (i, j) be at position (i in described high spectrum image to be detected, j) pixel, μ is the mean vector of background in described high spectrum image to be detected, Σ is the covariance matrix of background in described high spectrum image to be detected.
8. method according to claim 6, is characterized in that, described Threshold segmentation module comprises:
Obtain submodule, for obtaining confidence factor γ
lowerand γ
upper, and, the grey level histogram of described abnormality detection result;
First determines submodule, for the grey level histogram by described abnormality detection result, determines and described confidence factor γ
lowercorresponding gray threshold TH
lower, and with described confidence factor γ
uppercorresponding gray threshold TH
upper;
Second determines submodule, for from interval [TH
lower, TH
upper] in determine optimum gradation threshold value TH
best;
The 3rd determines submodule, for described abnormality detection result being greater than to described optimum gradation threshold value TH
bestpixel be defined as goal pels, obtain described goal pels collection, the pixel the goal pels except determining is defined as to backdrop pels, obtain described backdrop pels collection.
9. device according to claim 8, is characterized in that, described second determines that submodule comprises:
Interval division unit, for by described interval [TH
lower, TH
upper] be divided into a plurality of isometric intervals;
The first determining unit, for determining that described abnormality detection result falls into the quantity of each interval abnormality detection result value;
The second determining unit, for by the quantity that falls into each interval abnormality detection result value of determining, determines described optimum gradation threshold value TH by preset rules
best;
10. according to the device described in any one in claim 6-9, it is characterized in that, described the second detection module comprises:
The second calculating sub module, for utilizing mean vector and the covariance matrix of described backdrop pels collection, by
calculate the distance of described pixel to be detected and backdrop pels, wherein, μ
0for the mean vector of described backdrop pels collection, Σ
0for the covariance matrix of described backdrop pels collection, D
0(i, j) is the distance of described pixel r to be detected (i, j) and backdrop pels;
The 3rd calculating sub module, for utilizing mean vector and the covariance matrix of described goal pels collection, by
calculate the distance of described pixel to be detected and goal pels, wherein, r (i, j) is pixel to be detected, μ
1for the mean vector of described goal pels collection, Σ
1for the covariance matrix of described goal pels collection, D
1(i, j) is the distance of described pixel r to be detected (i, j) and goal pels.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310551785.7A CN103559715B (en) | 2013-11-07 | 2013-11-07 | A kind of method for detecting abnormality of high spectrum image and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310551785.7A CN103559715B (en) | 2013-11-07 | 2013-11-07 | A kind of method for detecting abnormality of high spectrum image and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103559715A true CN103559715A (en) | 2014-02-05 |
CN103559715B CN103559715B (en) | 2016-04-27 |
Family
ID=50013954
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310551785.7A Active CN103559715B (en) | 2013-11-07 | 2013-11-07 | A kind of method for detecting abnormality of high spectrum image and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103559715B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105427300A (en) * | 2015-12-21 | 2016-03-23 | 复旦大学 | Low-rank expression and learning dictionary-based hyperspectral image abnormity detection algorithm |
CN105989597A (en) * | 2015-02-13 | 2016-10-05 | 中国科学院西安光学精密机械研究所 | Hyperspectral image anomaly target detection method based on pixel selection process |
CN108399622A (en) * | 2018-03-01 | 2018-08-14 | 哈尔滨工业大学 | Hyperspectral image abnormal detection method based on non-negative sparse characteristic |
CN111818330A (en) * | 2020-07-02 | 2020-10-23 | 浙江大华技术股份有限公司 | Image abnormality detection method and apparatus, storage medium, and electronic apparatus |
CN112990313A (en) * | 2021-03-16 | 2021-06-18 | 中国科学院长春光学精密机械与物理研究所 | Hyperspectral image anomaly detection method and device, computer equipment and storage medium |
CN113269726A (en) * | 2021-04-29 | 2021-08-17 | 中国电子科技集团公司信息科学研究院 | Hyperspectral image target detection method and device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101144860A (en) * | 2007-10-16 | 2008-03-19 | 哈尔滨工业大学 | Hyperspectral image abnormal point detection method based on selective kernel principal component analysis |
US20110221926A1 (en) * | 2010-03-12 | 2011-09-15 | Kanaev Andrey V | Spectral Anomaly Detection in Deep Shadows |
CN102592280A (en) * | 2012-01-14 | 2012-07-18 | 哈尔滨工程大学 | Hyperspectral image anomaly detection method using multi-window feature analysis |
-
2013
- 2013-11-07 CN CN201310551785.7A patent/CN103559715B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101144860A (en) * | 2007-10-16 | 2008-03-19 | 哈尔滨工业大学 | Hyperspectral image abnormal point detection method based on selective kernel principal component analysis |
US20110221926A1 (en) * | 2010-03-12 | 2011-09-15 | Kanaev Andrey V | Spectral Anomaly Detection in Deep Shadows |
CN102592280A (en) * | 2012-01-14 | 2012-07-18 | 哈尔滨工程大学 | Hyperspectral image anomaly detection method using multi-window feature analysis |
Non-Patent Citations (4)
Title |
---|
QIANDONG GUO 等: "A target segmentation algorithm based on multivariate statistics and RX anomaly detection", 《ALGORITHMS AND TECHNOLOGIES FOR MULTISPECTRAL, HYPERSPECTRAL, AND ULTRASPECTRAL IMAGERY XVIII》 * |
XU SUN 等: "Feature-level fusion of multiple target detection results in hyperspectral image based on RX detector", 《AUTOMATIC TARGET RECOGNITION XXII》 * |
朱凤阳: "高光谱图像的异常目标检测及亚像元定位研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
赵春晖 等: "核加权RX高光谱图像异常检测算法", 《红外与毫米波学报》 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105989597A (en) * | 2015-02-13 | 2016-10-05 | 中国科学院西安光学精密机械研究所 | Hyperspectral image anomaly target detection method based on pixel selection process |
CN105989597B (en) * | 2015-02-13 | 2020-05-12 | 中国科学院西安光学精密机械研究所 | Hyperspectral image abnormal target detection method based on pixel selection process |
CN105427300A (en) * | 2015-12-21 | 2016-03-23 | 复旦大学 | Low-rank expression and learning dictionary-based hyperspectral image abnormity detection algorithm |
CN105427300B (en) * | 2015-12-21 | 2018-08-07 | 复旦大学 | A kind of hyperspectral image abnormal detection method based on low-rank representation and study dictionary |
CN108399622A (en) * | 2018-03-01 | 2018-08-14 | 哈尔滨工业大学 | Hyperspectral image abnormal detection method based on non-negative sparse characteristic |
CN108399622B (en) * | 2018-03-01 | 2021-04-09 | 哈尔滨工业大学 | Hyperspectral image anomaly detection method based on nonnegative sparse characteristics |
CN111818330A (en) * | 2020-07-02 | 2020-10-23 | 浙江大华技术股份有限公司 | Image abnormality detection method and apparatus, storage medium, and electronic apparatus |
CN112990313A (en) * | 2021-03-16 | 2021-06-18 | 中国科学院长春光学精密机械与物理研究所 | Hyperspectral image anomaly detection method and device, computer equipment and storage medium |
CN113269726A (en) * | 2021-04-29 | 2021-08-17 | 中国电子科技集团公司信息科学研究院 | Hyperspectral image target detection method and device |
Also Published As
Publication number | Publication date |
---|---|
CN103559715B (en) | 2016-04-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103559715B (en) | A kind of method for detecting abnormality of high spectrum image and device | |
EP2871610B1 (en) | Infrared image based early detection of oil spills in water | |
JP5501176B2 (en) | Ship detection apparatus, ship detection program, and ship detection method for ship detection apparatus | |
CN102509099B (en) | Detection method for image salient region | |
CN102542570B (en) | Method for automatically detecting dangerous object hidden by human body in microwave image | |
CN107742113B (en) | One kind being based on the posterior SAR image complex target detection method of destination number | |
CN102346910A (en) | Single-frame infrared image based real-time detection method of point target | |
CN102722706A (en) | Particle filter-based infrared small dim target detecting and tracking method and device | |
CN104680538A (en) | SAR image CFAR target detection method on basis of super pixels | |
CN103500453B (en) | Based on the SAR image salient region detection method of Gamma distribution and neighborhood information | |
CN103871039A (en) | Generation method for difference chart in SAR (Synthetic Aperture Radar) image change detection | |
CN104680553A (en) | Bilateral filtration-based variable kernel function target tracing method | |
CN101236646B (en) | Method and system for detecting distinct strong relevant direction with estimated image | |
Hwang et al. | A practical algorithm for the retrieval of floe size distribution of Arctic sea ice from high-resolution satellite Synthetic Aperture Radar imagery | |
CN103065299A (en) | Ultrasonic image edge extraction method and ultrasonic image edge extraction device | |
CN103065320A (en) | Synthetic aperture radar (SAR) image change detection method based on constant false alarm threshold value | |
CN103475800A (en) | Method and device for detecting foreground in image sequence | |
CN103559714A (en) | Anomaly detection method and device | |
CN110147716A (en) | Wave method for detecting area in a kind of SAR image combined based on frequency domain with airspace | |
CN106960433B (en) | It is a kind of that sonar image quality assessment method is referred to based on image entropy and the complete of edge | |
CN102360503A (en) | SAR (Specific Absorption Rate) image change detection method based on space approach degree and pixel similarity | |
CN106056115B (en) | A kind of infrared small target detection method under non-homogeneous background | |
CN103700109A (en) | Synthetic aperture radar (SAR) image change detection method based on multi-objective evolutionary algorithm based on decomposition (MOEA/D) and fuzzy clustering | |
CN106526545A (en) | Method for detecting robust CFAR detector | |
CN104881670B (en) | A kind of fast target extracting method for SAR orientation angular estimation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant |