CN103413136B - A kind of detection method of infrared image mesohigh line - Google Patents
A kind of detection method of infrared image mesohigh line Download PDFInfo
- Publication number
- CN103413136B CN103413136B CN201310332610.7A CN201310332610A CN103413136B CN 103413136 B CN103413136 B CN 103413136B CN 201310332610 A CN201310332610 A CN 201310332610A CN 103413136 B CN103413136 B CN 103413136B
- Authority
- CN
- China
- Prior art keywords
- voltage line
- pixel
- sigma
- infrared image
- gradient
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Image Analysis (AREA)
- Transforming Light Signals Into Electric Signals (AREA)
Abstract
The present invention relates to the detection method of a kind of infrared image mesohigh line, first the gradient of infrared image is calculated, carry out region growth, obtain the candidate region of high-voltage line, ridge detection method based on image second order derivative is used to obtain high-voltage line pixel in candidate region, high-voltage line pixel is attached and screens according to Freeman chain code, the high-voltage line finally detected.Compared with prior art, the present invention is not to the linear character using Hough transform to extract high-voltage line, but the classical lines detection methods such as region growth, ridge detection and Freeman chain code that fully utilize are to realize the detection of high-voltage line, the present invention can quickly detect the high-voltage line in infrared image, location high-voltage line can be arrived to sub-pixel precision, positioning precision is high, can effectively remove the impact of background noise in infrared image, foreign material, and robustness is good.
Description
Technical field
The invention belongs to infrared image processing technical field, be specifically related to the detection side of a kind of infrared image mesohigh line
Method.
Background technology
In the last few years, the using value of helicopter improves constantly, and function and range constantly expand.In the face of more and more multiple
Miscellaneous environment, nap of the earth flight has become various countries' aircraft conventional technique.And during helicopter super low altitude flight, the wire such as high-voltage line barrier
Thing is hindered but to be difficult to detect.It is reported, the reason that helicopter has an accident almost half be that collision high-pressure line barrier causes.
As can be seen here, the detection technique of research high-voltage line has critically important realistic meaning.
High-voltage line has the temperature characteristic apparently higher than surrounding, has brightness clearly special in infrared image
Levy.Therefore, utilizing infrared imaging device can reach to detect the purpose of high-voltage line, initially, foreign military force i.e. utilizes infrared imaging
Instrument is as high-voltage line detecting devices, but is not equipped with respective image processing system, but directly uses crew's naked eyes to see
Examine mode.Owing to human eye is limited to the observation ability of obstruction, operating distance is the nearest, it is impossible to provide for helicopter pilot
The enough response time.
Chinese Patent Application No. 201210376272.2, describes one and " automatically identifies aerial high-voltage from infrared image
The method of line ", core concept is to utilize random Hough transformation (RHT) method to realize detection to high-voltage line pixel, the method without
Method avoids Hough transform to calculate complexity, the shortcoming that parameter is many, processes the infrared image that background is complicated, easily can cause substantial amounts of
Flase drop.
Summary of the invention
Solve the technical problem that
In place of the deficiencies in the prior art, the present invention proposes the detection method of a kind of infrared image mesohigh line.
Technical scheme
The detection method of a kind of infrared image mesohigh line, it is characterised in that step is as follows:
Step 1: calculate gradient magnitude and the direction of infrared image;Direction according to gradient carries out region growth, obtains height
Line ball candidate region;
Step 2: in high-voltage line candidate region, uses ridge detection method based on image second order derivative, obtains high pressure
Line pixel;
Step 3: high-voltage line pixel be attached and screen according to Freeman chain code, finally being detected
High-voltage line.
Described step 1 carries out region and increases that to obtain the step of high-voltage line candidate region as follows:
Step a: use the gradient of Sobel operator 3 × 3 formwork calculation image each pixel of I, gradient magnitude and deflection
For: Wherein Gi,GjHorizontal and vertical component for pixel gradient G;
Step b: according to the size of each pixel gradient, the pixel p choosing Grad maximum carries out region as starting point
Increase, if starting pixel gradient direction angle is θpIf, neighborhood territory pixel gradient angle θ of pixel ppnWith θpAbsolute difference little
In 2 π/k(k=16), then this pixel is joined in the R of high-voltage line candidate region, the deflection in region is updated to simultaneouslyWhen repeating this process until not having new pixel to increase in region, this region R increases
Growth process terminates, and then starts next starting point, until infrared image does not has as the pixel of starting point, and to answer
Ensureing that all of pixel is at most accessed once, the pixel accessed can not be accessed again, will eventually get a series of
High-voltage line candidate region { R1,R2,...Rm}。
Described step 2 uses ridge detection method based on image second order derivative, obtains the step of high-voltage line pixel such as
Under:
Step a: to each pixel in infrared image I (x, y) carries out Gaussian convolution computing, obtain corresponding single order,
Second order directional is:
Ix=gσ(y)g'σ(x)*I(x,y)Iy=g'σ(y)gσ(x)*I(x,y)
Ixx=gσ(y)g''σ(x)*I(x,y)Ixy=g'σ(y)g'σ(x)*I(x,y)Iyy=g''σ(y)gσ(x)*I(x,y)
Wherein,
Step b: calculate each pixel in infrared image high-voltage line candidate region (x, y) corresponding Hessian matrix Eigenvalue, obtain maximum eigenvalue characteristic of correspondence vector (nx,ny).Along (nx,ny) direction
Calculate the location of pixels (p making first directional derivative be zerox,py):
If (px,py) ∈ ([-0.5,0.5], [-0.5,0.5]) and (x y) waits at same high-voltage line with pixel
In favored area, then this pixel is high-voltage line pixel.
The high-voltage line pixel of described step 3 is attached method: choose the high-voltage line picture that Second order directional is maximum
Vegetarian refreshments pstartFor the starting point of high-voltage line, neighborhood is found next high-voltage line pixel p aroundnextConnect into high-voltage line,
This pixel should make d+ β obtain minima, and wherein d represents the distance between the two high-voltage line pixel, and β represents the two
High-voltage line pixel (nx,ny) differential seat angle in direction.High-voltage line is connected into according to this rule recurrence.
The Freeman chain code screening of described step 3 is: if representing the chain of horizontal direction in high-voltage line Freeman chain code
Code accounts for the ratio of chain code sum and is less than 0.6, then remove this high-voltage line, otherwise retain this high-voltage line.Through screening, finally examined
The high-voltage line surveyed.
Beneficial effect
The detection method of a kind of infrared image mesohigh line that the present invention proposes, first calculates the gradient of infrared image, enters
Row region increases, and obtains the candidate region of high-voltage line, uses ridge detection side based on image second order derivative in candidate region
Method obtains high-voltage line pixel, is attached high-voltage line pixel and screens according to Freeman chain code, finally being examined
The high-voltage line surveyed.
Compared with prior art, employing Hough transform is not extracted the linear character of high-voltage line by the present invention, but
The classical lines detection methods such as region growth, ridge detection and Freeman chain code that fully utilize are to realize the inspection of high-voltage line
Surveying, the present invention can quickly detect the high-voltage line in infrared image, can arrive location high-voltage line to sub-pixel precision, positioning precision
Height, can effectively remove the impact of background noise in infrared image, foreign material, and robustness is good.
Accompanying drawing explanation
Fig. 1 is flow chart of the present invention;
Fig. 2 is infrared image;
Fig. 3 is the high-voltage line candidate region result figure after Fig. 2 carries out region growth;
Fig. 4 is the result figure of high-voltage line detection.
Detailed description of the invention
In conjunction with embodiment, accompanying drawing, the invention will be further described:
Embodiment of the present invention step is as follows:
Step 1: calculate the gradient of infrared image, carries out region growth and obtains high-voltage line candidate region, particularly as follows: initially with
The gradient of Sobel operator 3 × 3 formwork calculation infrared image each pixel of I, gradient magnitude and deflection be: Wherein Gi,GjHorizontal and vertical component for pixel gradient G;Then big according to each pixel gradient
Little, the pixel p choosing Grad maximum carries out region growth as starting point, if starting pixel gradient direction angle is θp, as
Really neighborhood territory pixel gradient angle θ of pixel ppnWith θpAbsolute difference less than 2 π/k(k=16), then this pixel is joined high pressure
In the R of line candidate region, the deflection in region is updated to simultaneouslyRepeat this process straight
When increasing in region to not having new pixel, this region R propagation process terminates, and then starts next starting point, until red
Outer image do not have can be as the pixel of starting point till, and should ensure that all of pixel is at most accessed once, accessed
Pixel can not be accessed again.Will eventually get a series of high-voltage line candidate region { R1,R2,...Rm}。
Step 2: in high-voltage line candidate region, uses ridge detection method based on image second order derivative, obtains high pressure
Line pixel, particularly as follows: first to each pixel in infrared image I, (x, y) carries out Gaussian convolution computing, obtains corresponding
Single order, Second order directional be:
Ix=gσ(y)g'σ(x)*I(x,y)Iy=g'σ(y)gσ(x)*I(x,y)
Ixx=gσ(y)g''σ(x)*I(x,y)Ixy=g'σ(y)g'σ(x)*I(x,y)Iyy=g''σ(y)gσ(x)*I(x,y)
Wherein,
Then each pixel in infrared image high-voltage line candidate region (x, y) corresponding Hessian matrix are calculated Eigenvalue, obtain maximum eigenvalue characteristic of correspondence vector (nx,ny).Along (nx,ny) direction
Calculate the location of pixels (p making first directional derivative be zerox,py):
Step 3: high-voltage line pixel is attached and screens according to Freeman chain code, it is achieved the essence of high-voltage line
First really detection and localization, particularly as follows: choose the high-voltage line pixel p that Second order directional is maximumstartFor the starting point of high-voltage line,
Neighborhood is found next high-voltage line pixel p aroundnextConnecting into high-voltage line, this pixel should make d+ β obtain minimum
Value, wherein d represents the distance between the two high-voltage line pixel, and β represents the two high-voltage line pixel (nx,ny) direction
Differential seat angle.Connect into high-voltage line according to this rule recurrence, during connecting, the high-voltage line connected is carried out simultaneously
Freeman chain code encodes.Then according to high-voltage line Freeman chain code, high-voltage line is screened, if high-voltage line Freeman chain
Represent the chain code of horizontal direction in Ma and account for the ratio of chain code sum less than 0.6, then remove this high-voltage line, otherwise retain this high pressure
Line.Through screening, the high-voltage line finally detected.
Specific embodiment:
Hardware environment for implementing is: Intel Duo 2 double-core 2.93G computer, 2.0GB internal memory, 512M video card, fortune
The software environment of row is: Microsoft Visual Studio2008, Windows XP.We use Microsoft Visual
Studio2008 software achieves the method that the present invention proposes.Infrared image size is 582*458.
1, the gradient of the infrared image in calculating Fig. 2, carries out region growth and obtains high-voltage line candidate region, particularly as follows: first
First using the gradient of Sobel operator 3 × 3 formwork calculation infrared image each pixel of I, gradient magnitude and deflection be: Wherein Gi,GjHorizontal and vertical component for pixel gradient G;Then according to each
The size of individual pixel gradient, the pixel p choosing Grad maximum carries out region growth as starting point, if starting pixel gradient
Deflection is θpIf, neighborhood territory pixel gradient angle θ of pixel ppnWith θpAbsolute difference less than 2 π/k(k=16), then should
Pixel joins in the R of high-voltage line candidate region, is updated to by the deflection in region simultaneously
When repeating this process until not having new pixel to increase in region, this region R propagation process terminates, and then starts the next one
Starting point, until infrared image does not has as the pixel of starting point, and to should ensure that all of pixel is the most interviewed
Asking once, the pixel accessed can not be accessed again.Will eventually get a series of high-voltage line candidate region { R1,R2,
...Rm, as shown in Figure 3.
2, in high-voltage line candidate region, use ridge detection method based on image second order derivative, obtain high-voltage line picture
Vegetarian refreshments, particularly as follows: first to each pixel in the infrared image in Fig. 1, (x y) carries out Gaussian convolution computing, obtains phase
Single order, the Second order directional answered be:
Ix=gσ(y)g'σ(x)*I(x,y)Iy=g'σ(y)gσ(x)*I(x,y)
Ixx=gσ(y)g''σ(x)*I(x,y)Ixy=g'σ(y)g'σ(x)*I(x,y)Iyy=g''σ(y)gσ(x)*I(x,y)
Wherein,
Then each pixel in infrared image high-voltage line candidate region (x, y) corresponding Hessian matrix are calculated Eigenvalue, obtain maximum eigenvalue characteristic of correspondence vector (nx,ny).Along (nx,ny) direction
Calculate the location of pixels (p making first directional derivative be zerox,py):
3, high-voltage line pixel it is attached and screens according to Freeman chain code, it is achieved accurately determining of high-voltage line
First position detection, particularly as follows: choose the high-voltage line pixel p that Second order directional is maximumstartFor the starting point of high-voltage line, in week
Enclose and neighborhood is found next high-voltage line pixel pnextConnecting into high-voltage line, this pixel should make d+ β obtain minima, its
Middle d represents the distance between the two high-voltage line pixel, and β represents the two high-voltage line pixel (nx,ny) angle in direction
Difference.Connect into high-voltage line according to this rule recurrence, during connecting, the high-voltage line connected is carried out Freeman chain simultaneously
Code coding.Then according to high-voltage line Freeman chain code, high-voltage line is screened, if high-voltage line Freeman chain code represents
The chain code of horizontal direction accounts for the ratio of chain code sum and is less than 0.6, then remove this high-voltage line, otherwise retain this high-voltage line.Through this mistake
After journey screening, obtain final high-voltage line testing result, as shown in Figure 4.
Claims (4)
1. the detection method of an infrared image mesohigh line, it is characterised in that step is as follows:
Step 1: calculate gradient magnitude and the direction of infrared image;Direction according to gradient carries out region growth, obtains high-voltage line
Candidate region;
Step 2: in high-voltage line candidate region, uses ridge detection method based on image second order derivative, obtains high-voltage line picture
Vegetarian refreshments;
Step 3: high-voltage line pixel is attached and screens according to Freeman chain code, the high pressure finally detected
Line;
Described step 2 uses ridge detection method based on image second order derivative, and the step obtaining high-voltage line pixel is as follows:
Step a: (x y) carries out Gaussian convolution computing, obtains corresponding single order, second order to each pixel in infrared image I
Directional derivative is:
Ix=gσ(y)g’σ(x) * I (x, y) Iy=g 'σ(y)gσ(x) * I (x, y)
Ixx=gσ(y)g″σ(x)*I(x,y) Ixy=g 'σ(y)g’σ(x)*I(x,y) Iyy=g "σ(y)gσ(x)*I(x,y)
Wherein,
Step b: calculate each pixel in infrared image high-voltage line candidate region (x, y) corresponding Hessian matrix Eigenvalue, obtain maximum eigenvalue characteristic of correspondence vector (nx,ny);Along (nx,ny) telegoniometer
Calculate the location of pixels (p making first directional derivative be zerox,py):
If (px,py) ∈ ([-0.5,0.5], [-0.5,0.5]) and (x, y) at same high-voltage line candidate regions with pixel
In territory, then this pixel is high-voltage line pixel.
The detection method of infrared image mesohigh line the most according to claim 1, it is characterised in that: described step 1 carries out district
The step that territory growth obtains high-voltage line candidate region is as follows:
Step a: using the gradient of Sobel operator 3 × 3 formwork calculation image each pixel of I, gradient magnitude and deflection be:Wherein Gi,GjHorizontal and vertical component for pixel gradient G;
Step b: according to the size of each pixel gradient, the pixel p choosing Grad maximum carries out region growth as starting point,
If starting pixel gradient direction angle is θpIf, neighborhood territory pixel gradient angle θ of pixel ppnWith θpAbsolute difference less than 2 π/
K, k=16, then join this pixel in the R of high-voltage line candidate region, be updated to by the deflection in region simultaneouslyWhen repeating this process until not having new pixel to increase in region, this region R increases
Growth process terminates, and then starts next starting point, until infrared image does not has as the pixel of starting point, and to answer
Ensureing that all of pixel is at most accessed once, the pixel accessed can not be accessed again, will eventually get a series of
High-voltage line candidate region { R1,R2,...Rm}。
The detection method of infrared image mesohigh line the most according to claim 1, it is characterised in that: the high pressure of described step 3
Line pixel is attached method: choose the high-voltage line pixel p that Second order directional is maximumstartInitiateing for high-voltage line
Point, finds next high-voltage line pixel p around in neighborhoodnextConnecting into high-voltage line, this pixel should make d+ β obtain
Little value, wherein d represents the distance between the two high-voltage line pixel, and β represents the two high-voltage line pixel (nx,ny) direction
Differential seat angle, according to this rule recurrence connect into high-voltage line.
The detection method of infrared image mesohigh line the most according to claim 1, it is characterised in that: described step 3
The screening of Freeman chain code is: if the ratio that the chain code representing horizontal direction in high-voltage line Freeman chain code accounts for chain code sum is little
In 0.6, then remove this high-voltage line, otherwise retain this high-voltage line, through screening, the high-voltage line finally detected.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310332610.7A CN103413136B (en) | 2013-08-01 | 2013-08-01 | A kind of detection method of infrared image mesohigh line |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310332610.7A CN103413136B (en) | 2013-08-01 | 2013-08-01 | A kind of detection method of infrared image mesohigh line |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103413136A CN103413136A (en) | 2013-11-27 |
CN103413136B true CN103413136B (en) | 2016-08-10 |
Family
ID=49606143
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310332610.7A Active CN103413136B (en) | 2013-08-01 | 2013-08-01 | A kind of detection method of infrared image mesohigh line |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103413136B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106203353B (en) * | 2016-07-14 | 2019-11-15 | 桂林长海发展有限责任公司 | A kind of detection system and method for undercarriage |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102842039A (en) * | 2012-07-11 | 2012-12-26 | 河海大学 | Road image detection method based on Sobel operator |
CN102930280A (en) * | 2012-10-05 | 2013-02-13 | 中国电子科技集团公司第十研究所 | Method for identifying overhead high-voltage wire automatically from infrared image |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8300949B2 (en) * | 2010-05-18 | 2012-10-30 | Sharp Laboratories Of America, Inc. | Edge detection technique having improved feature visibility |
-
2013
- 2013-08-01 CN CN201310332610.7A patent/CN103413136B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102842039A (en) * | 2012-07-11 | 2012-12-26 | 河海大学 | Road image detection method based on Sobel operator |
CN102930280A (en) * | 2012-10-05 | 2013-02-13 | 中国电子科技集团公司第十研究所 | Method for identifying overhead high-voltage wire automatically from infrared image |
Non-Patent Citations (2)
Title |
---|
红外高压线塔检测方法研究;秦名扬 等;《舰船电子工程》;20130620(第228期);第125-128页 * |
融合梯度特征的红外目标跟踪;余博 等;《西安工业大学学报》;20120525;第32卷(第5期);第355-360页 * |
Also Published As
Publication number | Publication date |
---|---|
CN103413136A (en) | 2013-11-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103761524B (en) | A kind of linear goal identification based on image and extracting method | |
US9734398B2 (en) | Method and apparatus for identifying object | |
CN107392963A (en) | A kind of imitative hawkeye moving target localization method for soft autonomous air refuelling | |
CN103884281B (en) | A kind of rover obstacle detection method based on initiating structure light | |
CN103198480B (en) | Based on the method for detecting change of remote sensing image of region and Kmeans cluster | |
CN105910583A (en) | Space debris fast detection and positioning method based on space-borne visible camera | |
CN104515980A (en) | Method and device for ground moving target indication based on InSAR (interferometric synthetic aperture radar) formation | |
CN108364279A (en) | Determine the method that stationary orbit remote sensing satellite is directed toward deviation | |
Kwak et al. | Registration of aerial imagery and aerial LiDAR data using centroids of plane roof surfaces as control information | |
CN105023279B (en) | A kind of video image space movement target detection method based on movable information | |
CN102999915B (en) | Meteorite crater matching method based on area ratio | |
CN104463821A (en) | Method for fusing infrared image and visible light image | |
CN103413136B (en) | A kind of detection method of infrared image mesohigh line | |
CN104299197A (en) | Single image defogging method by means of 2D CCA | |
CN106289319B (en) | A kind of satellite system biography letter measuring method based on in-orbit sequence image | |
CN104809720A (en) | Small cross view field-based double-camera target associating method | |
CN103606146B (en) | A kind of angular-point detection method based on circular target | |
CN104463896A (en) | Image corner point detection method and system based on kernel similar region distribution characteristics | |
Sturz et al. | Detecting the perception of illusory spatial boundaries: Evidence from distance judgments | |
CN106682668A (en) | Power transmission line geological disaster monitoring method using unmanned aerial vehicle to mark images | |
CN106056613A (en) | Magnetic resonance phase unwrapping method based on pixel classification and local surface fitting | |
CN106295699A (en) | A kind of earthquake Damage assessment method and apparatus based on high-definition remote sensing data | |
Krishnan et al. | Change detection using airborne lidar: Applications to earthquakes | |
Cao et al. | Extraction of water surfaces in simulated Ka-band SAR images of KaRIn on SWOT | |
CN103077525B (en) | Based on the method for detecting change of remote sensing image of Treelet image co-registration |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right |
Effective date of registration: 20180322 Address after: No. 1, Qin Du District, Shaanxi Province, Xianyang City, Shaanxi Patentee after: Shaanxi Zhongke Kai Hang Technology Co., Ltd. Address before: 710072 Xi'an friendship West Road, Shaanxi, No. 127 Patentee before: Northwestern Polytechnical University |
|
TR01 | Transfer of patent right |