CN108765335A - A kind of forest fire detection method based on remote sensing images - Google Patents
A kind of forest fire detection method based on remote sensing images Download PDFInfo
- Publication number
- CN108765335A CN108765335A CN201810512059.7A CN201810512059A CN108765335A CN 108765335 A CN108765335 A CN 108765335A CN 201810512059 A CN201810512059 A CN 201810512059A CN 108765335 A CN108765335 A CN 108765335A
- Authority
- CN
- China
- Prior art keywords
- image
- pixel
- binary
- structural elements
- obtains
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 30
- 230000007797 corrosion Effects 0.000 claims abstract description 27
- 238000005260 corrosion Methods 0.000 claims abstract description 27
- 230000009466 transformation Effects 0.000 claims abstract description 27
- 238000010586 diagram Methods 0.000 claims abstract description 10
- 238000001914 filtration Methods 0.000 claims abstract description 10
- 238000004364 calculation method Methods 0.000 claims description 27
- 238000000034 method Methods 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims 1
- BFWMWWXRWVJXSE-UHFFFAOYSA-M fentin hydroxide Chemical compound C=1C=CC=CC=1[Sn](C=1C=CC=CC=1)(O)C1=CC=CC=C1 BFWMWWXRWVJXSE-UHFFFAOYSA-M 0.000 claims 1
- 238000005516 engineering process Methods 0.000 abstract description 6
- 238000012544 monitoring process Methods 0.000 description 7
- 230000000877 morphologic effect Effects 0.000 description 5
- 230000011218 segmentation Effects 0.000 description 3
- 230000006378 damage Effects 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000005855 radiation Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000003331 infrared imaging Methods 0.000 description 1
- 238000002329 infrared spectrum Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 231100000241 scar Toxicity 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- 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
- G06T2207/10032—Satellite or aerial image; Remote sensing
-
- 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
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
Abstract
The forest fire detection method based on remote sensing images that the invention discloses a kind of being related to the intelligent identification technology field in machine vision commercial Application, and the present invention includes reading in initial pictures, carries out medium filtering and obtains denoising figure；It calculates local entropy and normalizes, obtain local entropy image；Morphology closed operation is carried out to local entropy diagram picture, then is corroded, corrosion image is obtained；Denoising figure carries out logarithmic transformation and obtains logarithmic transformation image；The gray level co-occurrence matrixes for calculating each pixel in logarithmic transformation image, obtain randomness, and randomness is normalized, obtain randomness normalized image；Two feature binary images are obtained respectively into row threshold division to corrosion image and randomness normalized image using maximum variance between clusters；Corresponding pixel in two feature binary images is carried out and operated respectively, obtains final output image, the present invention not only efficient quick, but also can ensure higher accuracy, there is practical value.
Description
Technical field
The present invention relates to the intelligent identification technology fields in machine vision commercial Application, are based on more particularly to one kind
The forest fire detection method of remote sensing images.
Background technology
Forest fire is a kind of global, the resource that jeopardizes forests disaster occurred every year, and forest is big each time
Fire all brings serious harm and loss to forest cover, forest ecosystem, global ecological environment and human life's property.China
It is heavy, extra big forest fires district occurred frequently has caused the great attention of government especially in northeast forest and south China forest.Therefore,
It is most important for the monitoring of forest fire, how to detect that forest fire has become the country accurately and in time
One of outer research hotspot.
Forest fire commonly takes place in unfrequented region, it is difficult to realize artificial detection.And satellite remote sensing technology exists
There is great advantage in terms of monitoring forest fire, remote sensing detection area is wide, and time, spatial resolution are high, of low cost, very
It is suitble to the relevant information extraction of fire, there is special ability and potentiality in forest fire detects work.From twentieth century seven
At the beginning of the ten's, domestic and foreign scholars' research forms the algorithm of a variety of detection forest fires in satellite infrared image, and is applied to
Fire monitoring, Burned scar mapping and the damage done by forest fires degree evaluation of Global Regional, such as based on forest fires infrared spectrum detection
MODIS fire point monitoring algorithms, the bright temperature threshold detection method based on imaging features of the forest fires on satellite infrared image and fire point
Etc..
It analyzed from remote sensing images, detect that forest fire is the key points and difficulties of remote sensing forest fire monitoring technology, it is existing
Some technologies focus principally on the detection of forest fire ignition point, and do not cover entire fire area, though the prior art
Fire can be so detected in time, but due to not having the morphological feature using forest fire infrared imaging, it is easy to by other height
The influence of radiation source or picture noise and lead to false-alarm；Moreover, not detected in first time on fire once, fire is spread apart
After no longer showing as ignition point, just it is likely to missing inspection occur；And it on the other hand, can not be true by existing detection method
The region of fire is determined with area.
Invention content
It is an object of the invention to：It is red using forest fire due to not having in order to solve existing forest fire detection method
The morphological feature of outer imaging, it is easy to the problem of influenced by other high radiation sources or picture noise and lead to false-alarm, the present invention
A kind of forest fire detection method based on remote sensing images is provided.
The present invention specifically uses following technical scheme to achieve the goals above：
A kind of forest fire detection method based on remote sensing images, includes the following steps：
S1, initial pictures f (x, y) is read in, carries out medium filtering and obtains denoising figure fpre(x, y), wherein (x, y) is indicated just
Pixel point coordinates in beginning image；
S2, denoising figure f is calculatedpre3 × 3 neighborhood local entropies and normalized of each pixel, obtain office in (x, y)
Portion's entropy diagram is as fent(x,y)；
S3, using 4 × 4 template to local entropy diagram as fent(x, y) carries out morphology closed operation, then with 3 × 3 templates pair
Local entropy image f after closed operationent(x, y) is corroded, and corrosion image f is obtainederode(x,y)；
S4, the denoising figure f that will be obtained in S1pre(x, y) carries out logarithmic transformation and obtains logarithmic transformation image flog(x,y)；
S5, logarithmic transformation image f is calculatedlog5 × 5 neighborhood gray level co-occurrence matrixes of each pixel in (x, y), further according to
Gray level co-occurrence matrixes obtain the randomness of each pixel, and the randomness of all pixels point is normalized, and obtain
Randomness normalized image frandom(x,y)；
S6, using maximum variance between clusters to corrosion image ferode(x, y) and randomness normalized image frandom(x,y)
Respectively into row threshold division, binary image f is obtainedbinary(x, y) and binary image f 'binary(x,y)；
S7, by binary image fbinary(x, y) and binary image f 'binaryCorresponding pixel difference in (x, y)
It carries out and operates, obtain final output image fout(x,y)。
Further, carrying out medium filtering to initial pictures f (x, y) in the S1 is specially：
3 × 3 neighborhood medium filterings are carried out to initial pictures f (x, y), obtain filtered denoising figure fpre(x, y), formula
For：
fm(x, y)=median { fround(x,y)}
fpre(x, y)=fm(x,y)
Wherein fround(x, y) indicates the gray value of each pixel in 3 × 3 neighborhoods, takes every in initial pictures f (x, y)
A pixel (x, y) calculates gray value f in each 3 × 3 neighborhood of pixel (x, y)roundThe intermediate value f of (x, y)m(x, y), then
With intermediate value fm(x, y) replaces the former ash angle value of corresponding pixel points (x, y), obtains denoising figure fpre(x,y)。
Further, the S2 specifically comprises the following steps：
S2.1, traversal denoising figure fpreIt is adjacent to count each pixel (x, y) 3 × 3 for each pixel (x, y) in (x, y)
The grey level histogram in domain, method are as follows：
The all pixels point for traversing each 3 × 3 neighborhood of pixel (x, y), if there are the gray values of pixel in neighborhood
For i, then the gray scale number h (i) of gray value i plus 1 after the completion of traversal, obtains gray probability p (i), formula is：
Wherein gray value i ∈ [0,255], M are denoising figure fpreThe height of (x, y), N are denoising figure fpreThe width of (x, y)；
S2.2, the local entropy H (x, y) for calculating each pixel (x, y), formula are：
H (x, y)=p (i) logp (i)；
S2.3, local entropy H (x, y) is normalized, obtains local entropy image fent(x, y), wherein local entropy image fent(x,
Y) arbitrary pixel (x in0,y0) calculation formula be：
Wherein min (H (x, y)) indicates that the minimum value in local entropy H (x, y), max (H (x, y)) indicate local entropy H (x, y)
In maximum value.
Further, the S3 specifically comprises the following steps：
S3.1, using 4 × 4 flat template as structural elements a, using structural elements a to local entropy diagram as fent(x, y) into
Row expansive working obtains expanding image fswell(x, y), calculation formula are：
Wherein fgray(x+s, y+t) belongs to when the center of structural elements a is at pixel (x, y), the figure that structural elements a is covered
As region；S, t expression make fgray(x+s, y+t) belongs to the constant in the region of structural elements a coverings；
S3.2, the inswept expanding image f of structural elements a are usedswellEach pixel of (x, y), obtains first time corrosion image
ferode1(x, y), calculation formula are：
Wherein fswell(x+u, y+v) belongs to when the center of structural elements a is at pixel (x, y), the figure that structural elements a is covered
As region；U, v expression make fswell(x+u, y+v) belongs to the constant in the region of structural elements a coverings；
S3.3, using 3 × 3 flat template as structural elements b, use the inswept first time corrosion image f of structural elements berode1
Each pixel of (x, y), obtains corrosion image ferode(x, y), calculation formula are：
Wherein ferode1(x+i, y+j) belongs to when the center of structural elements b is at pixel (x, y), and structural elements b is covered
Image-region；I, j expression make ferode1(x+i, y+j) belongs to the constant in the region of structural elements b coverings.
Further, the S4 specifically comprises the following steps：
S4.1, traversal denoising figure fpreEach pixel in (x, y) obtains preliminary log changing image flog1(x, y),
Calculation formula is：
flog1(x, y)=log (fpre(x,y)+1)；
S4.2, to preliminary log changing image flog1(x, y) normalized obtains logarithmic transformation image flog(x, y),
Wherein logarithmic transformation image flogArbitrary pixel (x in (x, y)0,y0) calculation formula be：
Wherein min (flog1(x, y)) indicate flog1Minimum value in (x, y), max (flog1(x, y)) indicate flog1(x,y)
In maximum value.
Further, the S5 specifically comprises the following steps：
If S5.1, logarithmic transformation image flogCurrent pixel point is (x in (x, y)0,y0), then its 5 × 5 neighborhood
Calculation formula is：
Wherein -2≤m≤2, -2≤n≤2；Traverse the neighborhood all pixels point, find out respectively the neighborhood 0 degree, 45 degree,
90 degree and 135 degree of gray level co-occurrence matrixes, the specific method is as follows：
S5.1.1, the gray level co-occurrence matrixes for enabling 0 degree, 45 degree, 90 degree and 135 degree are respectively g0(i,j)、g45(i,j)、g90
(i, j) and g135(i, j), the g0(i,j)、g45(i,j)、g90(i, j) and g135(i, j) is 255 × 255 matrix, traversal 5
All pixels point in × 5 neighborhoods, if preceding pixel point is (x ', y '),
IfThen g0(i, j)=g0(i,j)+1；
IfThen g45(i, j)=g45(i,j)+1；
IfThen g90(i, j)=g90(i,j)+1；
IfThen g135(i, j)=g135(i,j)+1；
S5.1.2, by gray level co-occurrence matrixes g0(i,j)、g45(i,j)、g90(i, j) and g135(i, j) is normalized,
Calculation formula is：
S5.2, according to the gray level co-occurrence matrixes after normalization, calculate current pixel point (x0,y0) at 0 degree, 45 degree, 90 degree and
135 degree of randomness r0(x0,y0)、r45(x0,y0)、r90(x0,y0) and r135(x0,y0), further according to randomness r0(x0,y0)、r45
(x0,y0)、r90(x0,y0) and r135(x0,y0) mean random is obtained to get to randomness normalized image frandom(x, y) when
Preceding pixel point (x0,y0) value frandom(x0,y0), calculation formula is：
Traverse logarithmic transformation image flogThe all pixels point of (x, y) is to get to randomness normalized image frandom(x,
y)。
Further, the S6 specifically comprises the following steps：
S6.1, corrosion image f is calculatederodeThe average gray of (x, y)；
S6.2, for gray value t (0≤t≤255), traversal corrosion image ferodeThe all pixels point of (x, y), according to every
All pixels point is divided into two parts by the gray value of a pixel, and a portion is pixel of the gray value less than or equal to t
Set A, another part are the set B of pixel of the gray value more than t；
S6.3, the ratio that the pixel number in set A and set B accounts for all pixels point number is calculated separately, is denoted as PA
And PB, then the average gray value of pixel in set A and set B is calculated separately, it is denoted asWith
S6.4, inter-class variance ICV is calculatedt, calculation formula is：
Enable t=1 successively, 2,3 ..., 255, compare to obtain maximum between-cluster varianceObtain binaryzation
Transform key t0；
S6.5, according to binaryzation transform key t0, by corrosion image ferode(x, y) is converted into binary image fbinary(x,
y)；Similarly, by randomness normalized image frandom(x, y) is converted into binary image f 'binary(x,y)。
Further, the S7 specifically comprises the following steps：
S7.1, traversal binary image fbinaryThe all pixels point of (x, y), current pixel point are (x0,y0), if
fbinary(x0,y0)=f 'binary(x0,y0)=1
Then fout(x0,y0)=1；
Otherwise fout(x0,y0)=0；
After the completion of traversal, final output image f is obtainedout(x,y)。
The present invention basic principle be：
The gray value shown in remote sensing images using forest fire is high and the lower spy of order of intensity profile
Point, to carry out the detection and segmentation of forest fire area, to analyze local entropy, the gray level co-occurrence matrixes characteristic value image of image,
In conjunction with the preprocess methods such as image median filter, logarithmic lengthening and Morphological scale-space, Threshold segmentation, vacancy filling
Equal later stages method for repairing and mending, the detection of forest fire and region segmentation in item remote sensing images.
Beneficial effects of the present invention are as follows：
1, the field local entropy and normalized and S5 that the present invention passes through each pixel in calculating denoising figure in S2
Each pixel neighborhood of a point gray level co-occurrence matrixes in middle calculating logarithmic transformation image, then the randomness of each pixel is calculated, and
The randomness of all pixels point is normalized, forest fire is extracted and is imaged unique morphological feature, it is not easy to by
To the influence of picture noise or other high radiating objects, Detection accuracy is improved, false drop rate is relatively low.
2, the present invention is carried out by corresponding pixel in binary image in S7 and binary image and is operated respectively,
Cascading judgement is carried out to target area using two kinds of features, can accurately detect and be partitioned into forest fire in remote sensing images
Region substantially increases accuracy in detection.
3, method and step efficient quick of the invention, processing time is short can be realized real in remote sensing forest fire monitoring
When forest fire detection.
Description of the drawings
Fig. 1 is the flow diagram of the present invention；
Fig. 2 is initial pictures f (x, y)；
Fig. 3 is corrosion image ferode(x,y)；
Fig. 4 is logarithmic transformation image flog(x,y)；
Fig. 5 is randomness normalized image frandom(x,y)；
Fig. 6 is binary image fbinary(x,y)；
Fig. 7 is binary image f 'binary(x,y)；
Fig. 8 is final output image fout(x,y)。
Specific implementation mode
In order to which those skilled in the art are better understood from the present invention, below in conjunction with the accompanying drawings with following embodiment to the present invention
It is described in further detail.
Embodiment 1
As shown in Figures 1 to 8, the forest fire detection method based on remote sensing images that the present embodiment provides a kind of, including such as
Lower step：
S1, initial pictures f (x, y) is read in, carries out medium filtering and obtains denoising figure fpre(x, y), wherein (x, y) is indicated just
Pixel point coordinates in beginning image；
S2, denoising figure f is calculatedpre3 × 3 neighborhood local entropies and normalized of each pixel, obtain office in (x, y)
Portion's entropy diagram is as fent(x,y)；
S3, using 4 × 4 template to local entropy diagram as fent(x, y) carries out morphology closed operation, then with 3 × 3 templates pair
Local entropy image f after closed operationent(x, y) is corroded, and corrosion image f is obtainederode(x,y)；
S4, the denoising figure f that will be obtained in S1pre(x, y) carries out logarithmic transformation and obtains logarithmic transformation image flog(x,y)；
S5, logarithmic transformation image f is calculatedlog5 × 5 neighborhood gray level co-occurrence matrixes of each pixel in (x, y), further according to
Gray level co-occurrence matrixes obtain the randomness of each pixel, and the randomness of all pixels point is normalized, and obtain
Randomness normalized image frandom(x,y)；
S6, using maximum variance between clusters to corrosion image ferode(x, y) and randomness normalized image frandom(x,y)
Respectively into row threshold division, binary image f is obtainedbinary(x, y) and binary image f 'binary(x,y)；
S7, by binary image fbinary(x, y) and binary image f 'binaryCorresponding pixel difference in (x, y)
It carries out and operates, obtain final output image fout(x,y)。
The field local entropy and normalized and S5 that the present embodiment passes through each pixel in calculating denoising figure in S2
Each pixel neighborhood of a point gray level co-occurrence matrixes in middle calculating logarithmic transformation image, then the randomness of each pixel is calculated, and
The randomness of all pixels point is normalized, forest fire is extracted and is imaged unique morphological feature, it is not easy to by
To the influence of picture noise or other high radiating objects, Detection accuracy is improved, false drop rate is relatively low；Pass through binary picture in S7
Corresponding pixel is carried out and is operated respectively in picture and binary image, combine sentencing to target area using two kinds of features
Certainly, the region that can accurately detect and be partitioned into forest fire in remote sensing images, substantially increases accuracy in detection；And this
The method and step efficient quick of embodiment, processing time is short can realize real-time Forest Fire in remote sensing forest fire monitoring
Calamity detects.
Embodiment 2
The present embodiment advanced optimizes on the basis of embodiment 1, specifically：
Carrying out medium filtering to initial pictures f (x, y) in the S1 is specially：
3 × 3 neighborhood medium filterings are carried out to initial pictures f (x, y), obtain filtered denoising figure fpre(x, y), formula
For：
fm(x, y)=median { fround(x,y)}
fpre(x, y)=fm(x,y)
Wherein fround(x, y) indicates the gray value of each pixel in 3 × 3 neighborhoods, takes every in initial pictures f (x, y)
A pixel (x, y) calculates gray value f in each 3 × 3 neighborhood of pixel (x, y)roundThe intermediate value f of (x, y)m(x, y), then
With intermediate value fm(x, y) replaces the former ash angle value of corresponding pixel points (x, y), obtains denoising figure fpre(x,y)。
The S2 specifically comprises the following steps：
S2.1, traversal denoising figure fpreIt is adjacent to count each pixel (x, y) 3 × 3 for each pixel (x, y) in (x, y)
The grey level histogram in domain, method are as follows：
The all pixels point for traversing each 3 × 3 neighborhood of pixel (x, y), if there are the gray values of pixel in neighborhood
For i, then the gray scale number h (i) of gray value i plus 1 after the completion of traversal, obtains gray probability p (i), formula is：
Wherein gray value i ∈ [0,255], M are denoising figure fpreThe height of (x, y), N are denoising figure fpreThe width of (x, y)；
S2.2, the local entropy H (x, y) for calculating each pixel (x, y), formula are：
H (x, y)=p (i) log p (i)；
S2.3, local entropy H (x, y) is normalized, obtains local entropy image fent(x, y), wherein local entropy image fent(x,
Y) arbitrary pixel (x in0,y0) calculation formula be：
Wherein min (H (x, y)) indicates that the minimum value in local entropy H (x, y), max (H (x, y)) indicate local entropy H (x, y)
In maximum value.
The S3 specifically comprises the following steps：
S3.1, using 4 × 4 flat template as structural elements a, using structural elements a to local entropy diagram as fent(x, y) into
Row expansive working obtains expanding image fswell(x, y), calculation formula are：
Wherein fgray(x+s, y+t) belongs to when the center of structural elements a is at pixel (x, y), the figure that structural elements a is covered
As region；S, t expression make fgray(x+s, y+t) belongs to the constant in the region of structural elements a coverings；
S3.2, the inswept expanding image f of structural elements a are usedswellEach pixel of (x, y), obtains first time corrosion image
ferode1(x, y), calculation formula are：
Wherein fswell(x+u, y+v) belongs to when the center of structural elements a is at pixel (x, y), the figure that structural elements a is covered
As region；U, v expression make fswell(x+u, y+v) belongs to the constant in the region of structural elements a coverings；
S3.3, using 3 × 3 templates as structural elements b, use the inswept first time corrosion image f of structural elements berode1(x,y)
Each pixel, obtain corrosion image ferode(x, y), calculation formula are：
Wherein ferode1(x+i, y+j) belongs to when the center of structural elements b is at pixel (x, y), and structural elements b is covered
Image-region；I, j expression make ferode1(x+i, y+j) belongs to the constant in the region of structural elements b coverings.
The S4 specifically comprises the following steps：
S4.1, traversal denoising figure fpreEach pixel in (x, y) obtains preliminary log changing image flog1(x, y),
Calculation formula is：
flog1(x, y)=log (fpre(x,y)+1)；
S4.2, to preliminary log changing image flog1(x, y) normalized obtains logarithmic transformation image flog(x, y),
Wherein logarithmic transformation image flogArbitrary pixel (x in (x, y)0,y0) calculation formula be：
Wherein min (flog1(x, y)) indicate flog1Minimum value in (x, y), max (flog1(x, y)) indicate flog1(x,y)
In maximum value.
The S5 specifically comprises the following steps：
If S5.1, logarithmic transformation image flogCurrent pixel point is (x in (x, y)0,y0), then its 5 × 5 neighborhood
Calculation formula is：
Wherein -2≤m≤2, -2≤n≤2；Traverse the neighborhood all pixels point, find out respectively the neighborhood 0 degree, 45 degree,
90 degree and 135 degree of gray level co-occurrence matrixes, the specific method is as follows：
S5.1.1, the gray level co-occurrence matrixes for enabling 0 degree, 45 degree, 90 degree and 135 degree are respectively g0(i,j)、g45(i,j)、g90
(i, j) and g135(i, j), the g0(i,j)、g45(i,j)、g90(i, j) and g135(i, j) is 255 × 255 matrix, traversal 5
All pixels point in × 5 neighborhoods, if preceding pixel point is (x ', y '),
IfThen g0(i, j)=g0(i,j)+1；
IfThen g45(i, j)=g45(i,j)+1；
IfThen g90(i, j)=g90(i,j)+1；
IfThen g135(i, j)=g135(i,j)+1；
S5.1.2, by gray level co-occurrence matrixes g0(i,j)、g45(i,j)、g90(i, j) and g135(i, j) is normalized,
Calculation formula is：
S5.2, according to the gray level co-occurrence matrixes after normalization, calculate current pixel point (x0,y0) at 0 degree, 45 degree, 90 degree and
135 degree of randomness r0(x0,y0)、r45(x0,y0)、r90(x0,y0) and r135(x0,y0), further according to randomness r0(x0,y0)、r45
(x0,y0)、r90(x0,y0) and r135(x0,y0) mean random is obtained to get to randomness normalized image frandom(x's, y)
Current pixel point (x0,y0) value frandom(x0,y0), calculation formula is：
Traverse logarithmic transformation image flogThe all pixels point of (x, y) is to get to randomness normalized image frandom(x,
y)。
The S6 specifically comprises the following steps：
S6.1, corrosion image f is calculatederodeThe average gray of (x, y)；
S6.2, for gray value t (0≤t≤255), traversal corrosion image ferodeThe all pixels point of (x, y), according to every
All pixels point is divided into two parts by the gray value of a pixel, and a portion is pixel of the gray value less than or equal to t
Set A, another part are the set B of pixel of the gray value more than t；
S6.3, the ratio that the pixel number in set A and set B accounts for all pixels point number is calculated separately, is denoted as PA
And PB, then the average gray value of pixel in set A and set B is calculated separately, it is denoted asWith
S6.4, inter-class variance ICV is calculatedt, calculation formula is：
Enable t=1 successively, 2,3 ..., 255, compare to obtain maximum between-cluster varianceObtain binaryzation
Transform key t0；
S6.5, according to binaryzation transform key t0, by corrosion image ferode(x, y) is converted into binary image fbinary(x,
y)；Similarly, by randomness normalized image frandom(x, y) is converted into binary image f 'binary(x,y)。
The S7 specifically comprises the following steps：
S7.1, traversal binary image fbinaryThe all pixels point of (x, y), current pixel point are (x0,y0), if
fbinary(x0,y0)=f 'binary(x0,y0)=1
Then fout(x0,y0)=1；
Otherwise fout(x0,y0)=0；
After the completion of traversal, final output image f is obtainedout(x,y)。
The above, only presently preferred embodiments of the present invention, are not intended to limit the invention, patent protection model of the invention
It encloses and is subject to claims, equivalent structure variation made by every specification and accompanying drawing content with the present invention, similarly
It should be included within the scope of the present invention.
Claims (8)
1. a kind of forest fire detection method based on remote sensing images, which is characterized in that include the following steps：
S1, initial pictures f (x, y) is read in, carries out medium filtering and obtains denoising figure fpre(x, y), wherein (x, y) indicates initial graph
Pixel point coordinates as in；
S2, denoising figure f is calculatedpre3 × 3 neighborhood local entropies and normalized of each pixel, obtain local entropy in (x, y)
Image fent(x,y)；
S3, using 4 × 4 template to local entropy diagram as fent(x, y) carries out morphology closed operation, then with 3 × 3 templates to closing behaviour
Local entropy image f after workent(x, y) is corroded, and corrosion image f is obtainederode(x,y)；
S4, the denoising figure f that will be obtained in S1pre(x, y) carries out logarithmic transformation and obtains logarithmic transformation image flog(x,y)；
S5, logarithmic transformation image f is calculatedlog5 × 5 neighborhood gray level co-occurrence matrixes of each pixel in (x, y), further according to gray scale
Co-occurrence matrix obtains the randomness of each pixel, and the randomness of all pixels point is normalized, and obtains random
Property normalized image frandom(x,y)；
S6, using maximum variance between clusters to corrosion image ferode(x, y) and randomness normalized image frandom(x, y) difference
Into row threshold division, binary image f is obtainedbinary(x, y) and binary image f 'binary(x,y)；
S7, by binary image fbinary(x, y) and binary image f 'binaryCorresponding pixel carries out respectively in (x, y)
With operation, final output image f is obtainedout(x,y)。
2. a kind of forest fire detection method based on remote sensing images according to claim 1, which is characterized in that the S1
In to initial pictures f (x, y) carry out medium filtering be specially：
3 × 3 neighborhood medium filterings are carried out to initial pictures f (x, y), obtain filtered denoising figure fpre(x, y), formula are：
fm(x, y)=median { fround(x,y)}
fpre(x, y)=fm(x,y)
Wherein fround(x, y) indicates the gray value of each pixel in 3 × 3 neighborhoods, takes each picture in initial pictures f (x, y)
Vegetarian refreshments (x, y) calculates gray value f in each 3 × 3 neighborhood of pixel (x, y)roundThe intermediate value f of (x, y)m(x, y), then with this
Intermediate value fm(x, y) replaces the former ash angle value of corresponding pixel points (x, y), obtains denoising figure fpre(x,y)。
3. a kind of forest fire detection method based on remote sensing images according to claim 1, which is characterized in that the S2
Specifically comprise the following steps：
S2.1, traversal denoising figure fpreEach pixel (x, y) in (x, y) counts each 3 × 3 neighborhood of pixel (x, y)
Grey level histogram, method are as follows：
Traverse all pixels point of each 3 × 3 neighborhood of pixel (x, y), if in neighborhood there are the gray value of pixel be i,
The then gray scale number h (i) of gray value i plus 1 after the completion of traversal, obtains gray probability p (i), formula is：
Wherein gray value i ∈ [0,255], M are denoising figure fpreThe height of (x, y), N are denoising figure fpreThe width of (x, y)；
S2.2, the local entropy H (x, y) for calculating each pixel (x, y), formula are：
H (x, y)=p (i) logp (i)；
S2.3, local entropy H (x, y) is normalized, obtains local entropy image fent(x, y), wherein local entropy image fentIn (x, y)
Arbitrary pixel (x0,y0) calculation formula be：
Wherein min (H (x, y)) indicates that the minimum value in local entropy H (x, y), max (H (x, y)) indicate in local entropy H (x, y)
Maximum value.
4. a kind of forest fire detection method based on remote sensing images according to claim 1, which is characterized in that the S3
Specifically comprise the following steps：
S3.1, using 4 × 4 flat template as structural elements a, using structural elements a to local entropy diagram as fent(x, y) carries out swollen
Swollen operation obtains expanding image fswell(x, y), calculation formula are：
Wherein fgray(x+s, y+t) belongs to when the center of structural elements a is at pixel (x, y), the image district that structural elements a is covered
Domain；S, t expression make fgray(x+s, y+t) belongs to the constant in the region of structural elements a coverings；
S3.2, the inswept expanding image f of structural elements a are usedswellEach pixel of (x, y) obtains first time corrosion image ferode1
(x, y), calculation formula are：
Wherein fswell(x+u, y+v) belongs to when the center of structural elements a is at pixel (x, y), the image district that structural elements a is covered
Domain；U, v expression make fswell(x+u, y+v) belongs to the constant in the region of structural elements a coverings；
S3.3, using 3 × 3 flat template as structural elements b, use the inswept first time corrosion image f of structural elements berode1(x,
Y) each pixel, obtains corrosion image ferode(x, y), calculation formula are：
Wherein ferode1(x+i, y+j) belongs to when the center of structural elements b is at pixel (x, y), the image that structural elements b is covered
Region；I, j expression make ferode1(x+i, y+j) belongs to the constant in the region of structural elements b coverings.
5. a kind of forest fire detection method based on remote sensing images according to claim 1, which is characterized in that the S4
Specifically comprise the following steps：
S4.1, traversal denoising figure fpreEach pixel in (x, y) obtains preliminary log changing image flog1(x, y) is calculated public
Formula is：
flog1(x, y)=log (fpre(x,y)+1)；
S4.2, to preliminary log changing image flog1(x, y) normalized obtains logarithmic transformation image flog(x, y), wherein right
Transformation of variables image flogArbitrary pixel (x in (x, y)0,y0) calculation formula be：
Wherein min (flog1(x, y)) indicate flog1Minimum value in (x, y), max (flog1(x, y)) indicate flog1In (x, y) most
Big value.
6. a kind of forest fire detection method based on remote sensing images according to claim 1, which is characterized in that the S5
Specifically comprise the following steps：
If S5.1, logarithmic transformation image flogCurrent pixel point is (x in (x, y)0,y0), then its 5 × 5 neighborhoodIt calculates
Formula is：
Wherein -2≤m≤2, -2≤n≤2；The neighborhood all pixels point is traversed, finds out the neighborhood respectively at 0 degree, 45 degree, 90 degree
With 135 degree of gray level co-occurrence matrixes, the specific method is as follows：
S5.1.1, the gray level co-occurrence matrixes for enabling 0 degree, 45 degree, 90 degree and 135 degree are respectively g0(i,j)、g45(i,j)、g90(i,j)
And g135(i, j), the g0(i,j)、g45(i,j)、g90(i, j) and g135(i, j) is 255 × 255 matrix, and traversal 5 × 5 is adjacent
All pixels point in domain, if preceding pixel point is (x ', y '),
IfThen g0(i, j)=g0(i,j)+1；
IfThen g45(i, j)=g45(i,j)+1；
IfThen g90(i, j)=g90(i,j)+1；
IfThen g135(i, j)=g135(i,j)+1；
S5.1.2, by gray level co-occurrence matrixes g0(i,j)、g45(i,j)、g90(i, j) and g135(i, j) is normalized, and calculates
Formula is：
S5.2, according to the gray level co-occurrence matrixes after normalization, calculate current pixel point (x0,y0) at 0 degree, 45 degree, 90 degree and 135
The randomness r of degree0(x0,y0)、r45(x0,y0)、r90(x0,y0) and r135(x0,y0), further according to randomness r0(x0,y0)、r45(x0,
y0)、r90(x0,y0) and r135(x0,y0) mean random is obtained to get to randomness normalized image frandom(x, y) current picture
Vegetarian refreshments (x0,y0) value frandom(x0,y0), calculation formula is：
Traverse logarithmic transformation image flogThe all pixels point of (x, y) is to get to randomness normalized image frandom(x,y)。
7. a kind of forest fire detection method based on remote sensing images according to claim 1, which is characterized in that the S6
Specifically comprise the following steps：
S6.1, corrosion image f is calculatederodeThe average gray of (x, y)
S6.2, for gray value t (0≤t≤255), traversal corrosion image ferodeThe all pixels point of (x, y), according to each picture
All pixels point is divided into two parts by the gray value of vegetarian refreshments, and a portion is the set of pixel of the gray value less than or equal to t
A, another part are the set B of pixel of the gray value more than t；
S6.3, the ratio that the pixel number in set A and set B accounts for all pixels point number is calculated separately, is denoted as PAAnd PB,
The average gray value for calculating separately pixel in set A and set B again, is denoted asWith
S6.4, inter-class variance ICV is calculatedt, calculation formula is：
Enable t=1 successively, 2,3 ..., 255, compare to obtain maximum between-cluster varianceObtain binaryzation conversion
Threshold value t0；
S6.5, according to binaryzation transform key t0, by corrosion image ferode(x, y) is converted into binary image fbinary(x,y)；
Similarly, by randomness normalized image frandom(x, y) is converted into binary image f 'binary(x,y)。
8. a kind of forest fire detection method based on remote sensing images according to claim 1, which is characterized in that the S7
Specifically comprise the following steps：
S7.1, traversal binary image fbinaryThe all pixels point of (x, y), current pixel point are (x0,y0), if
fbinary(x0,y0)=f 'binary(x0,y0)=1
Then fout(x0,y0)=1；
Otherwise fout(x0,y0)=0；
After the completion of traversal, final output image f is obtainedout(x,y)。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810512059.7A CN108765335B (en) | 2018-05-25 | 2018-05-25 | Forest fire detection method based on remote sensing image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810512059.7A CN108765335B (en) | 2018-05-25 | 2018-05-25 | Forest fire detection method based on remote sensing image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108765335A true CN108765335A (en) | 2018-11-06 |
| CN108765335B CN108765335B (en) | 2022-08-02 |
Family
ID=64006160
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810512059.7A Active CN108765335B (en) | 2018-05-25 | 2018-05-25 | Forest fire detection method based on remote sensing image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108765335B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109431511A (en) * | 2018-11-14 | 2019-03-08 | 南京航空航天大学 | A kind of human body back scoliosis angle measurement method based on Digital Image Processing |
| CN110119682A (en) * | 2019-04-04 | 2019-08-13 | 北京理工雷科电子信息技术有限公司 | A kind of infrared remote sensing Image Fire point recognition methods |
| CN110135269A (en) * | 2019-04-18 | 2019-08-16 | 杭州电子科技大学 | A kind of fire image detection method based on blend color model and neural network |
| CN111898681A (en) * | 2020-07-31 | 2020-11-06 | 北京林业大学 | Burned area fine extraction method based on remote sensing cloud platform and decision tree |
| CN112232318A (en) * | 2020-12-14 | 2021-01-15 | 北京华云星地通科技有限公司 | Cloud plus-end intelligent fire remote sensing monitoring and early warning method |
| CN113283276A (en) * | 2020-12-30 | 2021-08-20 | 四川弘和通讯有限公司 | Linkage thermal imaging self-learning fire point detection method and system |
| CN113628118A (en) * | 2020-05-06 | 2021-11-09 | 北京君正集成电路股份有限公司 | Denoising and filtering method for flat area |
| CN113822839A (en) * | 2020-06-18 | 2021-12-21 | 飞依诺科技(苏州)有限公司 | Medical image processing method and device, computer equipment and storage medium |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2001067415A1 (en) * | 2000-03-09 | 2001-09-13 | Robert Bosch Gmbh | Imaging fire detector |
| CN103886344A (en) * | 2014-04-14 | 2014-06-25 | 西安科技大学 | Image type fire flame identification method |
| CN103942806A (en) * | 2014-05-08 | 2014-07-23 | 西安电子科技大学 | Hyperspectral data based multi-information fusing ignition point detection method |
| WO2014179482A1 (en) * | 2013-04-30 | 2014-11-06 | The Regents Of The University Of California | Fire urgency estimator in geosynchronous orbit (fuego) |
| CN104318707A (en) * | 2014-11-04 | 2015-01-28 | 无锡北斗星通信息科技有限公司 | Forest fire detection platform based on satellite remote sensing images |
| CN104408706A (en) * | 2014-09-30 | 2015-03-11 | 天津艾思科尔科技有限公司 | A method for detecting and locating fire based on double-waveband image |
| CN104853151A (en) * | 2015-04-17 | 2015-08-19 | 张家港江苏科技大学产业技术研究院 | Large-space fire monitoring system based on video image |
| CN105046693A (en) * | 2015-06-29 | 2015-11-11 | 嘉兴慧康智能科技有限公司 | Eye fundus image microaneurysm automatic detection method |
| WO2016207875A1 (en) * | 2015-06-22 | 2016-12-29 | Photomyne Ltd. | System and method for detecting objects in an image |
| CN106355584A (en) * | 2016-08-30 | 2017-01-25 | 上海交通大学 | Automatic detection method for microaneurysm in eye fundus image on basis of local entropy determining threshold |
| CN106778582A (en) * | 2016-12-07 | 2017-05-31 | 哈尔滨工业大学 | Flame/smog recognition methods after forest map picture cutting based on RGB reconstruct |
| CN107256566A (en) * | 2017-05-26 | 2017-10-17 | 北京环境特性研究所 | Forest fires detection method based on emittance |
| CN107316012A (en) * | 2017-06-14 | 2017-11-03 | 华南理工大学 | The fire detection and tracking of small-sized depopulated helicopter |
| CN107705300A (en) * | 2017-09-28 | 2018-02-16 | 成都大熊智能科技有限责任公司 | A kind of method that blank page detection is realized based on morphological transformation |
-
2018
- 2018-05-25 CN CN201810512059.7A patent/CN108765335B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2001067415A1 (en) * | 2000-03-09 | 2001-09-13 | Robert Bosch Gmbh | Imaging fire detector |
| WO2014179482A1 (en) * | 2013-04-30 | 2014-11-06 | The Regents Of The University Of California | Fire urgency estimator in geosynchronous orbit (fuego) |
| CN103886344A (en) * | 2014-04-14 | 2014-06-25 | 西安科技大学 | Image type fire flame identification method |
| CN103942806A (en) * | 2014-05-08 | 2014-07-23 | 西安电子科技大学 | Hyperspectral data based multi-information fusing ignition point detection method |
| CN104408706A (en) * | 2014-09-30 | 2015-03-11 | 天津艾思科尔科技有限公司 | A method for detecting and locating fire based on double-waveband image |
| CN104318707A (en) * | 2014-11-04 | 2015-01-28 | 无锡北斗星通信息科技有限公司 | Forest fire detection platform based on satellite remote sensing images |
| CN104853151A (en) * | 2015-04-17 | 2015-08-19 | 张家港江苏科技大学产业技术研究院 | Large-space fire monitoring system based on video image |
| WO2016207875A1 (en) * | 2015-06-22 | 2016-12-29 | Photomyne Ltd. | System and method for detecting objects in an image |
| CN105046693A (en) * | 2015-06-29 | 2015-11-11 | 嘉兴慧康智能科技有限公司 | Eye fundus image microaneurysm automatic detection method |
| CN106355584A (en) * | 2016-08-30 | 2017-01-25 | 上海交通大学 | Automatic detection method for microaneurysm in eye fundus image on basis of local entropy determining threshold |
| CN106778582A (en) * | 2016-12-07 | 2017-05-31 | 哈尔滨工业大学 | Flame/smog recognition methods after forest map picture cutting based on RGB reconstruct |
| CN107256566A (en) * | 2017-05-26 | 2017-10-17 | 北京环境特性研究所 | Forest fires detection method based on emittance |
| CN107316012A (en) * | 2017-06-14 | 2017-11-03 | 华南理工大学 | The fire detection and tracking of small-sized depopulated helicopter |
| CN107705300A (en) * | 2017-09-28 | 2018-02-16 | 成都大熊智能科技有限责任公司 | A kind of method that blank page detection is realized based on morphological transformation |
Non-Patent Citations (3)
| Title |
|---|
| EDA, B等: "Computer-aided diagnosis of human brain tumor through MRI: A survey and a new algorithm", 《EXPERT SYSTEMS WITH APPLICATIONS》 * |
| 张宾: "基于视频的火灾火焰检测", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 王媛彬: "基于图像特征的煤矿火灾检测与识别研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109431511A (en) * | 2018-11-14 | 2019-03-08 | 南京航空航天大学 | A kind of human body back scoliosis angle measurement method based on Digital Image Processing |
| CN109431511B (en) * | 2018-11-14 | 2021-09-24 | 南京航空航天大学 | Human back scoliosis spine contour characteristic curve fitting method based on digital image processing |
| CN110119682A (en) * | 2019-04-04 | 2019-08-13 | 北京理工雷科电子信息技术有限公司 | A kind of infrared remote sensing Image Fire point recognition methods |
| CN110135269A (en) * | 2019-04-18 | 2019-08-16 | 杭州电子科技大学 | A kind of fire image detection method based on blend color model and neural network |
| CN110135269B (en) * | 2019-04-18 | 2020-12-08 | 杭州电子科技大学 | Fire image detection method based on mixed color model and neural network |
| CN113628118B (en) * | 2020-05-06 | 2023-12-08 | 北京君正集成电路股份有限公司 | Method for denoising and filtering in flat area |
| CN113628118A (en) * | 2020-05-06 | 2021-11-09 | 北京君正集成电路股份有限公司 | Denoising and filtering method for flat area |
| CN113822839A (en) * | 2020-06-18 | 2021-12-21 | 飞依诺科技(苏州)有限公司 | Medical image processing method and device, computer equipment and storage medium |
| CN113822839B (en) * | 2020-06-18 | 2024-01-23 | 飞依诺科技股份有限公司 | Medical image processing method, medical image processing device, computer equipment and storage medium |
| CN111898681A (en) * | 2020-07-31 | 2020-11-06 | 北京林业大学 | Burned area fine extraction method based on remote sensing cloud platform and decision tree |
| CN111898681B (en) * | 2020-07-31 | 2023-09-05 | 北京林业大学 | Fire trace fine extraction method based on remote sensing cloud platform and decision tree |
| CN112232318A (en) * | 2020-12-14 | 2021-01-15 | 北京华云星地通科技有限公司 | Cloud plus-end intelligent fire remote sensing monitoring and early warning method |
| CN113283276A (en) * | 2020-12-30 | 2021-08-20 | 四川弘和通讯有限公司 | Linkage thermal imaging self-learning fire point detection method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108765335B (en) | 2022-08-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108765335A (en) | A kind of forest fire detection method based on remote sensing images | |
| CN105067638B (en) | Tire fetal membrane face character defect inspection method based on machine vision | |
| CN105761261B (en) | A method of detection camera suffers artificial malicious sabotage | |
| CN105678213B (en) | Dual-mode mask person event automatic detection method based on video feature statistics | |
| Lawgaly et al. | Weighted averaging-based sensor pattern noise estimation for source camera identification | |
| CN109325959B (en) | Method for extracting infrared image details based on Hough transform and application thereof | |
| Sadaghiyanfam | Using gray-level-co-occurrence matrix and wavelet transform for textural fabric defect detection: A comparison study | |
| CN103353990A (en) | Intelligent-texture anti-counterfeiting method based on perceptual hashing | |
| CN103198319A (en) | Method of extraction of corner of blurred image in mine shaft environment | |
| CN109460705A (en) | Oil pipeline monitoring method based on machine vision | |
| CN114241370A (en) | Intrusion identification method and device based on digital twin transformer substation and computer equipment | |
| CN115171218A (en) | Material sample feeding abnormal behavior recognition system based on image recognition technology | |
| Joshi et al. | Damage identification and assessment using image processing on post-disaster satellite imagery | |
| CN111461003A (en) | Coal-fired working condition identification method based on video image sequence feature extraction | |
| CN112560574A (en) | River black water discharge detection method and recognition system applying same | |
| Saxena et al. | Video inpainting detection and localization using inconsistencies in optical flow | |
| CN113066041A (en) | Pavement crack detection method based on stack sparse self-coding deep learning | |
| CN101527041B (en) | Picture counterfeiting detection method based on shadow matte consistency | |
| CN108010076B (en) | End face appearance modeling method for intensive industrial bar image detection | |
| CN105718881B (en) | The zero illumination environment monitoring smoke dust method based on infrared video gray level image | |
| CN110555792B (en) | Image tampering blind detection method based on normalized histogram comprehensive feature vector | |
| JP6457728B2 (en) | Laminar smoke detection device and laminar smoke detection method | |
| Cao et al. | Infrared dim target detection via hand-crafted features and deep information combination | |
| Liu et al. | Splicing forgery localization via noise fingerprint incorporated with CFA configuration | |
| Jain et al. | An enhanced statistical approach for median filtering detection using difference image |
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 |