CN106683107A - SAR image edge detection method based on ROEWA improvement - Google Patents
SAR image edge detection method based on ROEWA improvement Download PDFInfo
- Publication number
- CN106683107A CN106683107A CN201610982408.2A CN201610982408A CN106683107A CN 106683107 A CN106683107 A CN 106683107A CN 201610982408 A CN201610982408 A CN 201610982408A CN 106683107 A CN106683107 A CN 106683107A
- Authority
- CN
- China
- Prior art keywords
- edge
- point
- roewa
- edge strength
- orientation
- 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.)
- Pending
Links
Classifications
-
- 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
- G06T2207/10044—Radar 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
- G06T2207/20036—Morphological image processing
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a SAR image edge detection method based on ROEWA improvement and relates to the computer image processing field. Through a ROEWA operator, an edge position and an edge direction can not be accurately determined. Aiming at the above problem, in the method of the invention, an expression is improved, and an inverted, signed and normalized minimum index weighting mean value ratio is redefined and is taken as an edge intensity index to quantitatively describe a jump degree of an edge. Based on that, an edge orientation is accurately calculated. Simultaneously, an improved non-maximum suppression algorithm is applied to an edge detection flow so that a detection effect of an existing SAR image edge detection technology scheme is greatly improved.
Description
Technical field
The present invention relates to Computer Image Processing field, specifically proposes a kind of based on the improved SAR image edges of ROEWA
Detection method.
Background technology
Rim detection is the basic problem in image procossing and computer vision, is to extract linear character and identification in image
The basis of linear goal in image.However, general Image Edge Detector is built based on Additive noise model, for tool
Synthetic aperture radar (SAR) image for having multiplicative noise does not have CFAR (Constant False Alarm) property, it may appear that
Retain false edge, omit the phenomenon of true edge.
Exponential weighting ratio of averages (ROEWA), is a kind of classical SAR image edge detection operator, and it is based on the property taken advantage of and makes an uproar
Acoustic model, the detection to SAR image has CFAR.But, the edge precision of ROEWA operators is poor, has to single edges
The phenomenon of many secondary responses, it is difficult to take into account the high detection rate and low false alarm rate at SAR image edge.
The domestic existing improved though based on ROEWA operators is:First with original ROEWA operators to SAR image
Carry out process and obtain edge strength, then according to edge strength using orientation estimator or orientation template estimation edge orientations,
Non-maxima suppression (Non-Maximum Suppression) is carried out finally according to edge orientations and edge strength to process.Non- pole
Big value Restrainable algorithms, the refinement of removal, edge in theory to false edge and are accurately positioned good effect, but need
There is provided accurate edge orientations could realize as parameter.At present, edge is concentrated mainly on based on the linguistic term of ROEWA
, there is the technical scheme of the orientation such as Gabor, Hough and Radon estimator in various orientation estimator aspects.
Whether the innovatory algorithm of orientation template is also based on based on orientation estimator, its edge orientations tried to achieve all can only
Edge is estimated on limited intended orientation.Usually, two improvements method be by circumference equal dividing be 4 orientation (i.e. 8 sides
To), shape " rice " word such as, every edge being located between 22.5 ° to 22.5 ° it is determined that 0 ° of (180 °) orientation, positioned at 22.5 °
Edge between 67.5 ° is it is determined that 45 ° of (225 °) orientation, other situations are by that analogy.Obviously, the edge side so tried to achieve
Position is inaccurate, and its possible value is limited and nor continuous distribution, so as to causing the effect of non-maxima suppression
It is barely satisfactory, fundamentally do not improve the defect of original ROEWA algorithms.
The content of the invention
Goal of the invention
The invention provides a kind of be based on the improved SAR image edge detection methods of ROEWA, SAR image edge is solved
The difficult problem that orientation is accurately calculated, and according to demand by non-maxima suppression algorithm improvement to sub-pixed mapping rank, edge precision
Well, high detection rate and low false alarm rate can be taken into account.
Technical scheme
To solve the above problems, the invention provides a kind of be based on the improved SAR image edge detection methods of ROEWA, bag
Include following steps:
(1) input gray level image;(2) rim detection is carried out with IROEWA operators, obtains edge strength figure and edge orientations;
(3) non-maxima suppression is carried out according to edge strength figure and edge orientations;(4) threshold value selection, binary segmentation and morphology are performed
Filtering Processing;(5) edge detection results figure is exported;
Wherein, IROEWA is that the edge strength weight expression to ROEWA is improved,
The edge strength weight expression of former ROEWA operators is:
The edge strength weight expression of IROEWA operators is:
Thus, according to formula D (x, y)=arctan [rY(x,y)/rX(x, y)], it is possible to accurately calculate (- pi/2, pi/2)
Interior azimuth, and judge, as azimuth D < 0, to make D ← D+ π so that azimuth codomain naturalization to [0, π) in;
Correspondingly, overall edge strength is then defined as:
The non-maxima suppression algorithm of intended orientation, the value for exactly judging current edge point is not in current edge orientation
It is local maximum:The big edge point sets such as space and edge strength figure are first built, using all picture dot points as undetermined
Marginal point;If current point is maximum after comparing with the neighborhood point on present orientation, protect current point as marginal point
Stay;If it is not, then current point is removed from edge point set.
Preferably, being changed to after IROEWA operators, calculated edge orientations have unlimited continuously possible value, when
For 0 °, 45 °, 90 °, 135 ° and 180 ° this special orientation when, neighborhood point is sub-pixed mapping, in addition it is also necessary to by the non-of intended orientation
Maximum Restrainable algorithms, are modified to the processing method of sub- picture dot rank.
Preferably, the improved method of sub-pixed mapping rank non-maxima suppression algorithm is:Given one can not be integer
Radius of neighbourhood r, and according to the edge orientations of current point, calculate the coordinate of two neighborhood points;If the transverse and longitudinal of neighborhood point is sat
Mark is not all integer, then need interpolation calculation its edge strength;It is downward upwards to the transverse and longitudinal coordinate of neighborhood point respectively during interpolation calculation
Round, combination of two can obtain the point of 4 rounded coordinates, then carry out linear interpolation according to this 4 points;Finally, current point is compared
With the edge strength size of neighborhood point, if the edge strength of current point is less than any one field point, by current point from side
Edge point is concentrated and rejected.
Beneficial effects of the present invention are as follows:The present invention cannot accurately determine marginal position and edge side for ROEWA operators
To problem, its expression formula is improved, redefined reversion, signed, normalized minimal index weighting
Average ratio, as edge strength index the transition degree at quantificational description edge is carried out, and accurately calculates edge on this basis
Orientation, while improved non-maxima suppression algorithm is applied in rim detection flow process, substantially improves existing SAR image side
The Detection results of edge detection technique scheme.
Description of the drawings
Fig. 1 is edge detecting technology route map;
Fig. 2 is the sub-pixed mapping level non-maxima suppression algorithm suitable for arbitrary orientation;
Specific embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present application, the technical scheme in the embodiment of the present application is carried out clear, complete
Site preparation is described, it is clear that described embodiment is only some embodiments of the present application, rather than the embodiment of whole.Based on this
Embodiment in application, the every other reality that those of ordinary skill in the art are obtained under the premise of creative work is not made
Example is applied, the scope of the application protection is belonged to.
By improved exponential weighting ratio of averages operator (Improved Ratio of Exponentially Weighted
Averages) IROEWA is referred to as, and the algorithm with sub-pixed mapping rank non-maxima suppression is referred to as IROEWA&NMS.Such as Fig. 1 institutes
Show, step of the present invention is as follows:(1) input gray level image;(2) rim detection is carried out with IROEWA operators, obtains edge strength figure
And edge orientations;(3) non-maxima suppression is carried out according to edge strength figure and edge orientations;(4) threshold value selection, two-value point are performed
Cut and the post processing such as morphologic filtering;(5) edge detection results figure is exported.Wherein, improved core link is (2nd) step.
IROEWA is the same with ROEWA, is to be based on linear minimum mean-squared error wave filter (Linear MMSE Filter),
Under one-dimensional case, wave filter f expression formulas are:
F (x)=Ce-α|x| (1)
In the discrete case, wave filter f can use causal filter f1With non-causal filter f2To realize, then wave filter f
Discrete representation is:
In formula, f1(n)=abnH (n), f2(n)=ab-nh(-n).And wherein, 0 < b=e-α< 1;H (n) is
Heaviside functions, in the discrete case:When n >=0, h (n)=1;Otherwise, h (n)=0.
Wave filter f is generalized to by two-dimensional space according to f (x, y)=f (x) f (y).So, for level orientation, first use
Wave filter f is filtered by column to image intensity I, and f is then used respectively1And f2Filtered line by line, you can respectively obtain cause and effect and
Non-causal exponential weighting average:
In formula, * represents level orientation convolution algorithm, and ⊙ represents vertical orientation convolution algorithm.In the same manner, vertical orientation can be obtained
Cause and effect and non-causal exponential weighting average.In formula (3) and (4), input signal and f1And f2Convolution algorithm can be simplified to recurrence
Computing.Assume wave filter f1And f2Input signal is e1And e2, then convolution s1And s2Recursion is as follows:
s1(n)=a [e1(n)-s1(n-1)]+s1(n-1), n=1 ..., N (5)
s2(n)=a [e2(n)-s2(n+1)]+s2(n+1), n=N ..., 1 (6)
Horizontally and vertically four kinds of exponential weighting averages in orientation can be tried to achieve by above-mentioned recursive algorithm association type (2).
IROEWA is exactly that the edge strength weight expression to ROEWA is improved.
The edge strength weight expression of former ROEWA operators is:
The edge strength weight expression of IROEWA operators is:
Thus, according to formula D (x, y)=arctan [rY(x,y)/rX(x, y)], it is possible to accurately calculate (- pi/2, pi/2)
Interior azimuth, and judge, as azimuth D < 0, to make D ← D+ π so that azimuth codomain naturalization to [0, π) in.
Correspondingly, overall edge strength is then defined as:
The non-maxima suppression algorithm of intended orientation, the value for exactly judging current edge point is not in current edge orientation
It is local maximum:The big edge point sets such as space and edge strength figure are first built, using all picture dot points as undetermined
Marginal point;If current point is maximum after comparing with the neighborhood point on present orientation, protect current point as marginal point
Stay;If it is not, then current point is removed from edge point set.And be changed to after IROEWA operators, calculated edge side
Position have unlimited continuously may value, when not being 0 °, 45 °, 90 °, 135 ° and 180 ° etc. this special orientation, neighborhood point
It is sub-pixed mapping.So, in addition it is also necessary to by the non-maxima suppression algorithm of intended orientation, it is modified to the processing method of sub- picture dot rank.
As shown in Fig. 2 the improved though of sub-pixed mapping rank non-maxima suppression algorithm is:Given one can not be integer
Radius of neighbourhood r, and according to the edge orientations of current point, calculate the coordinate of two neighborhood points;If the transverse and longitudinal of neighborhood point is sat
Mark is not all integer, then need interpolation calculation its edge strength;It is downward upwards to the transverse and longitudinal coordinate of neighborhood point respectively during interpolation calculation
Round, combination of two can obtain the point of 4 rounded coordinates, then carry out linear interpolation according to this 4 points;Finally, current point is compared
With the edge strength size of neighborhood point, if the edge strength of current point is less than any one field point, by current point from side
Edge point is concentrated and rejected.
Claims (3)
1. it is a kind of to be based on the improved SAR image edge detection methods of ROEWA, it is characterised in that to comprise the following steps:
(1) input gray level image;(2) rim detection is carried out with IROEWA operators, obtains edge strength figure and edge orientations;(3)
Non-maxima suppression is carried out according to edge strength figure and edge orientations;(4) threshold value selection, binary segmentation and morphologic filtering are performed
Process;(5) edge detection results figure is exported;
Wherein, IROEWA is that the edge strength weight expression to ROEWA is improved,
The edge strength weight expression of former ROEWA operators is:
(7)
The edge strength weight expression of IROEWA operators is:
Thus, according to formula D (x, y)=arctan [rY(x,y)/rX(x, y)], it is possible to accurately calculate in (- pi/2, pi/2)
Azimuth, and judge, as azimuth D < 0, to make D ← D+ π so that azimuth codomain naturalization to [0, π) in;
Correspondingly, overall edge strength is then defined as:
The non-maxima suppression algorithm of intended orientation, the value for exactly judging current edge point is office in current edge orientation
Portion's maximum:The big edge point sets such as a space and edge strength figure are first built, using all picture dot points as side undetermined
Edge point;If current point is maximum after comparing with the neighborhood point on present orientation, retain current point as marginal point;Such as
Fruit is not then to remove current point from edge point set.
2. it is according to claim 1 a kind of based on the improved SAR image edge detection methods of ROEWA, it is characterised in that to change
After for IROEWA operators, calculated edge orientations have unlimited continuously may value, when not being 0 °, 45 °, 90 °, 135 °
And during 180 ° of this special orientation, neighborhood point is sub-pixed mapping, in addition it is also necessary to by the non-maxima suppression algorithm of intended orientation, improve
Into the processing method of sub- picture dot rank.
3. it is according to claim 2 a kind of based on the improved SAR image edge detection methods of ROEWA, it is characterised in that sub-
The improved method of pixel rank non-maxima suppression algorithm is:Given one can not be integer radius of neighbourhood r, and according to working as
The edge orientations of front point, calculate the coordinate of two neighborhood points;If the transverse and longitudinal coordinate of neighborhood point is not all integer, interpolation is needed
Calculate its edge strength;During interpolation calculation, the transverse and longitudinal coordinate of neighborhood point is rounded downwards upwards respectively, combination of two can obtain 4
The point of rounded coordinate, then carries out linear interpolation according to this 4 points;Finally, current point is compared big with the edge strength of neighborhood point
It is little, if current point is concentrated and rejected by the edge strength of current point less than any one field point from marginal point.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610982408.2A CN106683107A (en) | 2016-11-08 | 2016-11-08 | SAR image edge detection method based on ROEWA improvement |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610982408.2A CN106683107A (en) | 2016-11-08 | 2016-11-08 | SAR image edge detection method based on ROEWA improvement |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106683107A true CN106683107A (en) | 2017-05-17 |
Family
ID=58840272
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610982408.2A Pending CN106683107A (en) | 2016-11-08 | 2016-11-08 | SAR image edge detection method based on ROEWA improvement |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106683107A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107564027A (en) * | 2017-10-25 | 2018-01-09 | 中国电子科技集团公司第五十四研究所 | A kind of edge vectors computational methods of diameter radar image |
CN110390338A (en) * | 2019-07-10 | 2019-10-29 | 武汉大学 | A kind of SAR high-precision matching process based on non-linear guiding filtering and ratio gradient |
CN110782471A (en) * | 2019-10-16 | 2020-02-11 | 中国矿业大学 | Multi-scale SAR image edge detection method |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105512665A (en) * | 2015-12-11 | 2016-04-20 | 中国测绘科学研究院 | Airborne laser radar point cloud data edge extraction method |
-
2016
- 2016-11-08 CN CN201610982408.2A patent/CN106683107A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105512665A (en) * | 2015-12-11 | 2016-04-20 | 中国测绘科学研究院 | Airborne laser radar point cloud data edge extraction method |
Non-Patent Citations (4)
Title |
---|
WU YIQUAN ET AL.: "Edge Detection of River in SAR Image Based on Contourlet Modulus Maxima and Improved Mathematical Morphology", 《TRANSACTIONS OF NANJING UNIVERSITY OF AERONAUTICS AND ASTRONAUTICS》 * |
刘夯 等: "SAR图像ROEWA边缘检测器的改进", 《遥感学报》 * |
安成锦 等: "基于改进ROEWA算子的SAR图像边缘检测方法", 《中国图象图形学报》 * |
贾承丽 等: "一种改进的SAR图像边缘检测方法", 《电子与信息学报》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107564027A (en) * | 2017-10-25 | 2018-01-09 | 中国电子科技集团公司第五十四研究所 | A kind of edge vectors computational methods of diameter radar image |
CN110390338A (en) * | 2019-07-10 | 2019-10-29 | 武汉大学 | A kind of SAR high-precision matching process based on non-linear guiding filtering and ratio gradient |
CN110390338B (en) * | 2019-07-10 | 2022-08-05 | 武汉大学 | SAR high-precision matching method based on nonlinear guided filtering and ratio gradient |
CN110782471A (en) * | 2019-10-16 | 2020-02-11 | 中国矿业大学 | Multi-scale SAR image edge detection method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP3076367A1 (en) | Method for road detection from one image | |
CN109635656A (en) | Vehicle attribute recognition methods, device, equipment and medium neural network based | |
CN103942758A (en) | Dark channel prior image dehazing method based on multiscale fusion | |
CN104732536A (en) | Sub-pixel edge detection method based on improved morphology | |
CN106683107A (en) | SAR image edge detection method based on ROEWA improvement | |
CN103778613B (en) | A kind of adaptive polarization SAR images filter method of window | |
CN117197140B (en) | Irregular metal buckle forming detection method based on machine vision | |
CN104038792B (en) | For the video content analysis method and apparatus of IPTV supervision | |
Yu et al. | Edge detection of agricultural products based on morphologically improved canny algorithm | |
CN115797872A (en) | Machine vision-based packaging defect identification method, system, equipment and medium | |
CN105844337A (en) | Intelligent garbage classification device | |
CN105678262A (en) | Tunnel geological monitoring equipment capable of conducting autonomous operation | |
CN109101976A (en) | A kind of detection method of arc extinguishing grid pieces surface defect | |
CN105741281A (en) | Image edge detection method based on neighbourhood dispersion | |
CN112417955A (en) | Patrol video stream processing method and device | |
CN111489389A (en) | Light spot center detection method | |
CN114187267A (en) | Stamping part defect detection method based on machine vision | |
Chen et al. | An enhanced adaptive Sobel edge detector based on improved genetic algorithm and non-maximum suppression | |
CN105414774A (en) | Laser cutting device capable of achieving autonomous cutting | |
CN105928099A (en) | Intelligent air purifier | |
CN101727668B (en) | Method and device for detecting image boundary | |
CN107169440A (en) | A kind of Approach for road detection based on graph model | |
CN116071415A (en) | Stereo matching method based on improved Census algorithm | |
CN115482239A (en) | Image positioning method, system, device and medium | |
CN101719273A (en) | On-line self-adaptation extraction method of metallurgy strip surface defect based on one-dimension information entropy |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170517 |