CN104156956B - A kind of multicorner edge detection operator method recognized based on Gauss wavelet one-dimensional peak value - Google Patents
A kind of multicorner edge detection operator method recognized based on Gauss wavelet one-dimensional peak value Download PDFInfo
- Publication number
- CN104156956B CN104156956B CN201410382235.1A CN201410382235A CN104156956B CN 104156956 B CN104156956 B CN 104156956B CN 201410382235 A CN201410382235 A CN 201410382235A CN 104156956 B CN104156956 B CN 104156956B
- Authority
- CN
- China
- Prior art keywords
- image
- rim detection
- dimensional
- edge
- row
- 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
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000003708 edge detection Methods 0.000 title claims abstract description 37
- 238000001514 detection method Methods 0.000 claims abstract description 56
- 239000013598 vector Substances 0.000 claims abstract description 49
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 18
- 238000004458 analytical method Methods 0.000 claims abstract description 5
- 238000006243 chemical reaction Methods 0.000 claims abstract description 4
- 230000017105 transposition Effects 0.000 claims description 7
- 210000003127 knee Anatomy 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 2
- 238000003709 image segmentation Methods 0.000 abstract description 3
- 230000008569 process Effects 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 9
- 230000008859 change Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 210000003414 extremity Anatomy 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000012067 mathematical method Methods 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000009394 selective breeding Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of multicorner edge detection operator method recognized based on Gauss wavelet one-dimensional peak value, which comprises the following steps：Histogram analysis are carried out to pending image, the gray value of target and its background is obtained, assignment is carried out to the 18 kinds of template images for giving using the gray value for obtaining；Preset some groups of rim detection parameters, using default some groups of rim detection parameters respectively to assignment after 18 kinds of template images carry out rim detection, and find out and template image degree of approximation highest edge detection results, rim detection parameter used by the result will be obtained as optimized parameter, using the optimized parameter for obtaining, by making following process to pending image：Image segmentation, some one-dimensional vectors built, is made convolution algorithm and is taken absolute value, determining local maximum, gray value is assigned to local maximum and being replaced the pixel in original image in same position and binary conversion treatment is carried out to the edge of image after multiple stacking, obtaining the image after rim detection.
Description
Technical field
The present invention relates to a kind of image processing method, especially with regard to a kind of based on many of Gauss wavelet one-dimensional peak value identification
Angled edge detection method.
Background technology
Picture strip gives people the vivid thinking world of class one, is the important channel in the human knowledge world.It is present in image
Mutation and discontinuously jiggly structure we term it edge.Edge often carries abundant image information.These sides
Edge point constitutes contour of object, and these profiles be often researcher interested where, it embodies a concentrated reflection of the feature of goal in research,
Have particularly important place mat effect to follow-up image segmentation, images match, target recognition, computer vision, thus how
In real image, structural unstable edge is refined into constitutionally stable edge becomes the side that people for many years further investigate always
To.In the research of decades, people continually introduce mathematical method and image border are extracted and is explained.By initial based on
The Prewitt operators of gradient principle, Sobel operators wait until LoG operators and Canny operators, then to wavelet transformation, embody edge
The depth and difficulty of test problems.
Multicorner edge detection operator method generally comprises template Operator Method and two-dimensional wavelet transformation mould based on gradient principle
Maximum edge detection method.Template Operator Method based on gradient principle is carried out to two dimensional image using the gradient template of N*N
Convolution.As template is generally square, and its size is 5 pixel *, 5 pixel to the maximum, so the gradient direction that template can be generated
Up to 16, i.e., 0 °, 30 °, 45 °, 60 °, 90 °, 120 °, 135 °, 150 °, 180 °, 210 °, 225 °, 240 °, 270 °, 300 °,
315 ° and 330 ° of directions.Classical two-dimensional wavelet transformation modulus maximum edge detection method can only seek ladder in the x-direction with y directions
Sorting out further according to argument after degree carries out non-maxima suppression.Image border is entered hence with existing multicorner edge detection operator method
The basic mode by rotation image and rotational coordinates of row detection is realized.But all can be to image when rotation image and rotational coordinates
Interpolation arithmetic is carried out, the change of gradation of image information is caused, so as to enter to the edge of image after rotation image and rotational coordinates
Cannot ensure the accuracy of image border during row identification, and also need to side after limb recognition is carried out according to the angle of rotation
Edge image rotation returns original position, so causes again the change of edge image information.In addition, rotation image and rotational coordinates meeting
Cause the change of picture size and produce image boundary problem, these problems can all increase image procossing difficulty.
The selection of edge detection threshold is to reach the efficient frontier for failing to detect to draw noise with mistake according to statistical decision
The point for rising is classified as of marginal point and trades off.The method that existing threshold value is selected includes artificial selection's method and automatic threshold method.Manually
Selection method is the simplest, can determine image border by eye-observation result, multiple interactive operation.But human eye sense in image
The details for feeling insensitive is possible to be ignored.It is to be sentenced according to certain decision method threshold value automatically to automatically select threshold method
Disconnected.It is applied to major applications environment, but finds suitable decision strategy and evaluate its effectiveness and have difficulties.With regard to side
The selection of edge detection algorithm parameter there is presently no unified standard, substantially determine calculation as standard whether to meet application
Method parameter.
Content of the invention
For the problems referred to above, it is an object of the invention to provide one kind can solve list on the premise of image information is not changed
The multicorner edge detection operator method recognized based on Gauss wavelet one-dimensional peak value of pixel multicorner edge detection operator problem.
For achieving the above object, the present invention takes technical scheme below：A kind of based on Gauss wavelet one-dimensional peak value recognize
Multicorner edge detection operator method, which comprises the following steps：1) histogram analysis are carried out to pending image, obtains target and its back of the body
The 18 kinds of template images for giving are carried out assignment using the gray value for obtaining by the gray value of scape；2) some groups of rim detection are preset
Parameter, using default some groups of rim detection parameters respectively to assignment after 18 kinds of template images carry out rim detection, and look for
Go out and template image degree of approximation highest edge detection results, the rim detection parameter obtained used by the result is joined as optimum
Number, which specifically includes following steps：(1) template image is split using the image subblock of m rows × mtan θ row, by template
Image segmentation is the image subblock of several m rows × mtan θ row, and wherein, θ is rim detection angle, θ ∈ [45 °, 90 °)；(2)
Some one-dimensional vectors are built along its diagonal in the image subblock of m rows × mtan θ row；(3) some one-dimensional vectors that will be built
First derivative f with Gaussian function respectivelyσT () makees convolution algorithm, and convolution algorithm result is taken absolute value；(4) determine convolution
As a result the local maximum of absolute value；(5) local maximum to obtaining assigns gray value, and the side that will be generated according to pixel subscript
Edge image pixel replaces the pixel in original image in same position；(6) some sides that different edge detection angles directions are obtained
Edge image carries out gray scale superposition, according to the grey level histogram of image after multiple stacking, gray percentage highest gray value is made
For binary-state threshold, binary conversion treatment is carried out to the edge of image after multiple stacking according to the binary-state threshold；(7) adopt
The image border of generation is compared by Pratt rim detection quality factor methods with image border in template image, obtain with
The template image edge degree of approximation one group of edge detection results of highest, and the rim detection used by the edge detection results will be obtained
Parameter is used as optimized parameter；3) adopt step 2) optimized parameter that obtains, repeat step 2) in the step of (1)～step
(6) rim detection is carried out to pending image,.
The step 2) in, default rim detection parameter includes first derivative f of Gaussian functionσParameter σ of (t), one
Dimension peak value maximum threshold alpha and binary-state threshold.
The step 2) in, one-dimensional vector is built to the image subblock of m rows × mtan θ row, which includes：1. examined according to edge
The span of measuring angle θ：θ ∈ [45 °, 90 °), provide three kinds of forms of the image subblock of m rows × mtan θ rowWith
Q90°,And Q90°It is expressed as with pixel a respectively：
θ1Form for image subblockCorresponding rim detection angle,
θ2Form for image subblockCorresponding rim detection angle,
2. for the form of the image subblock of m rows × mtan θ rowAccording to following form
Form along image subblockDiagonal take some one-dimensional vectors X1、X2…XmAnd Y1…Ym-1,
Form for the image subblock of m rows × mtan θ rowAccording to following form
Form along image subblockDiagonal take some one-dimensional vector X '1、X'2…X'mWith Y '1…Y′m-1,
Form Q for the image subblock of m rows × mtan θ row90°, one-dimensional vector X of structure "1、X″2…X″mFor：
3., after by image array transposition and upset, adopt and build with the image array sub-block corresponding to [45 °, 90 °]
One-dimensional vector identical method, builds one-dimensional vector to the image array sub-block corresponding to [0 °, 45 °]；By by image array
After upset, one-dimensional vector identical method is built using with the image array sub-block corresponding to [45 °, 90 °], to [90 °, 135 °]
Corresponding image array sub-block builds one-dimensional vector；After by image array transposition, using with corresponding to [45 °, 90 °]
Image array sub-block builds one-dimensional vector identical method, the image array sub-block corresponding to [135 °, 180 °] is built one-dimensional
Vector；Complete the structure of the one-dimensional vector to the image array sub-block corresponding to [0 °, 180 °].
The step 2) in, determine convolution algorithm result absolute value | fσ(t) * X | local maximum, which includes following step
Suddenly：1. preset a maximum threshold alpha, α=(0,1)；2. exist | fσ(t) * X | at first crest in waveform, before comparing successively
The size of data afterwards, when data are switched to successively decrease interval by incremental interval, records knee value MAX | fσ(t)*X|；When crest is passed
Subtract one data of interval presence and be less than α MAX | fσ(t) * X | when, then the crest is effectively and its maximum is MAX | fσ(t)*X|；If should
Descending tops interval is not present less than α MAX | fσ(t) * X | value and proceed to the incremental interval of next crest, then the crest without
Effect, continues the maximum for judging next crest；3. non-maximum is suppressed.
Due to taking above technical scheme, which has advantages below to the present invention：1st, the present invention is due to according to rim detection angle
Degree [45 °, 90 °) tangent value, select the image subblock of m rows × mtan θ row to split template image, by template image
Be divided into the image subblock of several m rows × mtan θ row, then some one-dimensional vectors are built along the diagonal of the image subblock；With
Based on 45 °～90 ° direction rim detection, by template image matrix is turned down and/or transposition after, realize to 0 °～
The structure of the one-dimensional vector of the image array sub-block corresponding to 180 °, so that reduce the angular range for building one-dimensional vector；Logical
Crossing carries out convolution algorithm and right by the one-dimensional vector of the image array sub-block corresponding to 0 °～180 ° and Gaussian function first derivative
Convolution results take absolute value, and make the angular range of edge direction detection expand to [0 °, 360 °] from [0 °, 180 °], therefore this
Bright using this one-dimensional vector segmentation figure picture by the way of expand the scope of rim detection angle, it is possible to achieve in angular interval
The rim detection of line number m of selected image subblock any one angle direction corresponding such that it is able to do not changing image information
On the premise of solve the problems, such as single pixel multicorner edge detection operator.2nd, the present invention due to will build one-dimensional vector respectively with Gaussian function
Several first derivatives makees convolution algorithm, and convolution algorithm result is taken absolute value, exhausted by finding each convolution algorithm result
To the local maximum being worth, two dimensional image rim detection problem is made to be changed into one-dimensional waveform peakvalue's checking problem, therefore the present invention
Image Edge-Detection difficulty can be reduced.3rd, the present invention is due to adopting default rim detection parameter to 18 kinds of given Prototype drawings
As carrying out rim detection, by comparing the image border and image border in template of generation, the rim detection parameter of optimum is obtained,
And rim detection is carried out to pending image using the optimum rim detection parameter, the edge for therefore being obtained using the present invention is examined
Survey result more accurate.Above advantage is based on, the present invention is can be widely applied in Image Edge-Detection.
Description of the drawings
Fig. 1 is the flow chart of the multicorner edge detection operator method that the present invention is recognized based on Gauss wavelet one-dimensional peak value；
Fig. 2 is given 18 kind template image；Wherein, figure (a) is the template image of 9 kinds of step edges, and figure (b) is 9 kinds
The template image at line edge；
Fig. 3 is the schematic diagram that template image or pending image is split using image subblock；Wherein, " ... " table
As the border of block, " " represents that the border of template image or pending image, θ represent rim detection angle to diagram, and 0 represents
0 pixel filling area；
Fig. 4 is to make convolution algorithm, take absolute value and peakvalue's checking effect diagram；
Fig. 5 is to determine the schematic diagram of the local maximum of convolution algorithm result absolute value；Wherein, figure (a) is Valid peak
Schematic diagram, figure (b) is invalid crest schematic diagram；
Fig. 6 is that 18 kinds of template images compare schematic diagram with its edge detection results；Wherein, figure (a) is the mould of step edges
Plate image compares schematic diagram with its edge detection results, schemes the template image that (b) is line edge and compares with its edge detection results
Schematic diagram.
Specific embodiment
With reference to the accompanying drawings and examples the present invention is described in detail.
As shown in figure 1, the present invention included based on the multicorner edge detection operator method that Gauss wavelet one-dimensional peak value is recognized following
Step：
1) histogram analysis are carried out to pending image, obtains the gray value of target and its background, using the gray scale for obtaining
Value carries out assignment to the 18 kinds of template images for giving, and which specifically includes：
(1) optionally piece image is used as pending image, and carries out histogram analysis to which, obtains mesh in pending image
Mark and its gray value of background.
(2) as shown in Fig. 2 giving 18 kinds of template images.
(3) gray value obtained using step (1) is respectively to the target and its background in 18 kinds of template images in step (2)
Carry out assignment.
2) some groups of rim detection parameters are preset：First derivative f of Gaussian functionσT parameter σ of (), one-dimensional peak value are very big
Value threshold alpha and binary-state threshold, using default some groups of rim detection parameters respectively to assignment after 18 kinds of template images enter
Row rim detection, and find out and template image degree of approximation highest edge detection results, the edge inspection used by the result will be obtained
Parameter is surveyed as optimized parameter.
Only as a example by carrying out rim detection using one group of rim detection parameter to 18 kinds of template images, illustrate to find optimum ginseng
Several processes, which specifically includes following steps：
(1) template image is split using image subblock；
As shown in figure 3, according to rim detection angle, θ ∈ [45 °, 90 °) tangent value, select the figure of m rows × mtan θ row
Template image is split as sub-block and using the image subblock, template image is divided into several m rows × mtan θ row
Image subblock.When template image can not be divided into the image subblock of integer m rows × mtan θ row, template image is expanded
Exhibition, the expansion of template image are filled up with 0 pixel grey scale, enable template image to be divided into integer m rows × mtan θ
The image subblock of row.
(2) diagonal along the image subblock in the image subblock of m rows × mtan θ row builds some one-dimensional vectors；
Method due to building one-dimensional vector in each image subblock is identical, therefore only with wherein m rows × mtan θ row
As a example by image subblock, construction method is illustrated, which specifically includes following steps：
1. according to the span of rim detection angle, θ：θ ∈ [45 °, 90 °), provide the image subblock of m rows × mtan θ row
Three kinds of formsAnd Q90°,And Q90°It is expressed as with pixel a respectively：
θ1Form for image subblockCorresponding rim detection angle,
θ2Form for image subblockCorresponding rim detection angle,
WithThe form one closed interval [θ of corresponding rim detection angle composition of two kinds of image subblocks1,θ2].Its angle
Scope is：
In formula, m ＞ 1.
2. for the form of the image subblock of m rows × mtan θ rowAccording to following form
Form along image subblockDiagonal take some one-dimensional vectors X1、X2…XmAnd Y1…Ym-1,
Form for the image subblock of m rows × mtan θ rowAccording to following form
Form along image subblockDiagonal take some one-dimensional vector X '1、X'2…X'mWith Y '1…Y′m-1,
Form Q for the image subblock of m rows × mtan θ row90°, one-dimensional vector X of structure "1、X″2…X″mFor：
3., after by image array transposition and upset, adopt and build with the image array sub-block corresponding to [45 °, 90 °]
One-dimensional vector identical method, builds one-dimensional vector to the image array sub-block corresponding to [0 °, 45 °].By by image array
After upset, one-dimensional vector identical method is built using with the image array sub-block corresponding to [45 °, 90 °], to [90 °, 135 °]
Corresponding image array sub-block builds one-dimensional vector.After by image array transposition, using with corresponding to [45 °, 90 °]
Image array sub-block builds one-dimensional vector identical method, the image array sub-block corresponding to [135 °, 180 °] is built one-dimensional
Vector.Above method is based on, and according to the method for one-dimensional vector being built to the image array sub-block corresponding to [45 °, 90 °], is realized
Structure to the one-dimensional vector of the image array sub-block corresponding to [0 °, 180 °] such that it is able to which reduction builds the angle of one-dimensional vector
Degree scope.
(3) as shown in figure 4, some one-dimensional vectors X that will be built1、X2…XmAnd Y1…Ym-1With Gaussian function one respectively
Order derivative fσT () makees convolution algorithm, and convolution algorithm result is taken absolute value obtain：
|fσ(t)*X1|,|fσ(t)*X2|,…|fσ(t)*Xm| and | fσ(t)*Y1|,…|fσ(t)*Ym-1| (4) so as to
Make in convolution algorithm result occur negative value be changed on the occasion of.
Make convolution and the computing for taking absolute value by some one-dimensional vectors to building, make rim detection angle from [0 °,
180 °] expand to [0 °, 360 °].Therefore at only need to be to the one-dimensional vector of the image array sub-block corresponding to [0 °, 180 °]
Reason.
(4) by the absolute value of each convolution algorithm result of searching | fσ(t) * X | and | fσ(t) * Y | local maximum,
Two dimensional image rim detection problem is made to be changed into one-dimensional waveform peakvalue's checking problem.Wherein, X=X1, X2..., Xm, Y=Y1,
Y2..., Ym-1.
As shown in figure 5, determining the absolute value of convolution algorithm result using following methods | fσ(t) * X | local maximum,
Which specifically includes following steps：
1. preset a maximum threshold alpha, α=(0,1).
2. exist | fσ(t) * X | at first crest in waveform, compare the size of before and after's data successively, when data are by being incremented by
Interval switch to successively decrease interval when, record knee value MAX | fσ(t)*X|；There are data less than α MAX when descending tops is interval |
fσ(t) * X | when, then the crest is effectively and its maximum is MAX | fσ(t)*X|；If the descending tops interval is not present less than α MAX
|fσ(t) * X | value and proceed to the incremental interval of next crest, then the crest is invalid, continues to judge the very big of next crest
Value.
3. non-maximum is suppressed.
(5) maximum for step (4) being obtained assigns gray value, and its value is：
According to pixel subscript, the edge image pixel for generating is replaced the pixel in original image in same position.
(6) some edge images that different edge detection angles directions obtain are carried out gray scale superposition.True edge position
Can brighten because of the multiple stacking of pixel grey scale.According to the grey level histogram of image after multiple stacking, by gray percentage highest
Gray value carries out binary conversion treatment according to the binary-state threshold to the edge of image after multiple stacking as binary-state threshold, raw
Into image border final as shown in Figure 6.
(7) Pratt rim detection quality factor method is adopted by the image border for generating and image border in template image
It is compared, obtains and the image border degree of approximation one group of edge detection results of highest in template image, and the side will be obtained
Rim detection parameter used by edge testing result is used as optimized parameter.
3) adopt step 2) optimized parameter that obtains, repeat step 2) in the step of (1)～step (6), to pending
Image carries out rim detection.
The various embodiments described above are merely to illustrate the present invention, the structure of wherein each part, connected mode and method and step etc. all
Can be varied from, every equivalents carried out on the basis of technical solution of the present invention and improvement should not be excluded
Outside protection scope of the present invention.
Claims (4)
1. a kind of based on Gauss wavelet one-dimensional peak value recognize multicorner edge detection operator method, which comprises the following steps：
1) histogram analysis are carried out to pending image, obtains the gray value of target and its background, using the gray value pair for obtaining
18 kinds of given template images carry out assignment；
2) preset some groups of rim detection parameters, using default some groups of rim detection parameters respectively to assignment after 18 kinds of moulds
Plate image carries out rim detection, and finds out and template image degree of approximation highest edge detection results, will obtain used by the result
Rim detection parameter as optimized parameter, which specifically includes following steps：
(1) template image is split using the image subblock of m rows × mtan θ row, template image is divided into several m rows
× mtan θ row image subblock, wherein, θ be rim detection angle, θ ∈ [45 °, 90 °)；
(2) some one-dimensional vectors are built along its diagonal in the image subblock of m rows × mtan θ row；
(3) first derivative f by some one-dimensional vectors for building respectively with Gaussian functionσT () makees convolution algorithm, and convolution is transported
Calculate result to take absolute value；
(4) determine the local maximum of convolution results absolute value；
(5) local maximum to obtaining assigns gray value, and the edge image pixel for generating is replaced artwork according to pixel subscript
Pixel as in same position；
(6) some edge images that different edge detection angles directions obtain are carried out gray scale superposition, is schemed according to after multiple stacking
The grey level histogram of picture, using gray percentage highest gray value as binary-state threshold, according to the binary-state threshold to multiple
After superposition, the edge of image carries out binary conversion treatment；
(7) image border of generation is carried out with image border in template image using Pratt rim detection quality factor methods
Relatively, obtain and the template image edge degree of approximation one group of edge detection results of highest, and the edge detection results institute will be obtained
Rim detection parameter is used as optimized parameter；
3) adopt step 2) optimized parameter that obtains, repeat step 2) in step (1)～step (6), pending image is entered
Row rim detection.
2. a kind of multicorner edge detection operator method recognized based on Gauss wavelet one-dimensional peak value as claimed in claim 1, which is special
Levy and be：The step 2) in, default rim detection parameter includes first derivative f of Gaussian functionσParameter σ of (t), one-dimensional
Peak value maximum threshold alpha and binary-state threshold.
3. as claimed in claim 1 or 2 a kind of based on Gauss wavelet one-dimensional peak value recognize multicorner edge detection operator method, its
It is characterised by：The step 2) in, one-dimensional vector is built to the image subblock of m rows × mtan θ row, which includes：
1. according to the span of rim detection angle, θ：θ ∈ [45 °, 90 °), provide the image subblock of m rows × mtan θ row
Three kinds of formsWithAnd Q90°It is expressed as with pixel a respectively：
θ1Form for image subblockCorresponding rim detection angle,
θ2Form for image subblockCorresponding rim detection angle,
2. for the form of the image subblock of m rows × mtan θ rowAccording to following form
Form along image subblockDiagonal take some one-dimensional vectors X1、X2…XmAnd Y1…Ym-1,
Form for the image subblock of m rows × mtan θ rowAccording to following form
Form along image subblockDiagonal take some one-dimensional vector X '1、X'2…X'mWith Y '1…Y′m-1,
Form Q for the image subblock of m rows × m tan θ row90°, one-dimensional vector X of structure "1、X″2…X″mFor：
3. after by image array transposition and upset, one-dimensional using building with the image array sub-block corresponding to [45 °, 90 °]
Vectorial identical method, builds one-dimensional vector to the image array sub-block corresponding to [0 °, 45 °]；By image array is overturn
Afterwards, one-dimensional vector identical method is built using with the image array sub-block corresponding to [45 °, 90 °], right to [90 °, 135 °] institutes
The image array sub-block that answers builds one-dimensional vector；After by image array transposition, adopt and the image corresponding to [45 °, 90 °]
Matrix sub block builds one-dimensional vector identical method, builds one-dimensional vector to the image array sub-block corresponding to [135 °, 180 °]；
Complete the structure of the one-dimensional vector to the image array sub-block corresponding to [0 °, 180 °].
4. as claimed in claim 1 or 2 a kind of based on Gauss wavelet one-dimensional peak value recognize multicorner edge detection operator method, its
It is characterised by：The step 2) in, determine convolution algorithm result absolute value | fσ(t) * X | local maximum, which includes following
Step：
1. preset a maximum threshold alpha, α=(0,1)；
2. exist | fσ(t) * X | at first crest in waveform, compare the size of before and after's data successively, when data are by incremental interval
Switch to successively decrease interval when, record knee value MAX | fσ(t)*X|；There are data less than α MAX when descending tops is interval | fσ(t)*X
| when, then the crest is effectively and its maximum is MAX | fσ(t)*X|；If the descending tops interval is not present less than α MAX | fσ(t)*
X | value and proceed to the incremental interval of next crest, then the crest is invalid, continues the maximum for judging next crest；
3. non-maximum is suppressed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410382235.1A CN104156956B (en) | 2014-08-06 | 2014-08-06 | A kind of multicorner edge detection operator method recognized based on Gauss wavelet one-dimensional peak value |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410382235.1A CN104156956B (en) | 2014-08-06 | 2014-08-06 | A kind of multicorner edge detection operator method recognized based on Gauss wavelet one-dimensional peak value |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104156956A CN104156956A (en) | 2014-11-19 |
| CN104156956B true CN104156956B (en) | 2017-03-15 |
Family
ID=51882446
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410382235.1A Active CN104156956B (en) | 2014-08-06 | 2014-08-06 | A kind of multicorner edge detection operator method recognized based on Gauss wavelet one-dimensional peak value |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104156956B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019041590A1 (en) * | 2017-08-31 | 2019-03-07 | 中国科学院微电子研究所 | Edge detection method using arbitrary angle |
| CN108320019B (en) * | 2018-02-06 | 2020-11-24 | 澎峰（北京）科技有限公司 | Convolution calculation method and device for deep convolution neural network |
| JP6813004B2 (en) * | 2018-06-28 | 2021-01-13 | Ｊｆｅスチール株式会社 | Steel non-pressure lower width detector and its detection method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103310461A (en) * | 2013-06-27 | 2013-09-18 | 清华大学深圳研究生院 | Image edge extraction method based on blocked Kalman filtering |
| CN103308056A (en) * | 2013-05-23 | 2013-09-18 | 中国科学院自动化研究所 | Road marking detection method |
-
2014
- 2014-08-06 CN CN201410382235.1A patent/CN104156956B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103308056A (en) * | 2013-05-23 | 2013-09-18 | 中国科学院自动化研究所 | Road marking detection method |
| CN103310461A (en) * | 2013-06-27 | 2013-09-18 | 清华大学深圳研究生院 | Image edge extraction method based on blocked Kalman filtering |
Non-Patent Citations (5)
| Title |
|---|
| Comparison between Various Edge Detection Methods on Satellite Image;H.S.Bhadauria等;《International Journal of Emerging Technology and Advanced Engineering》;20130630;第3卷(第6期);第324-328页 * |
| Edge and line oriented contour detection: State of the art;Giuseppe Papari等;《Image and Vision Computing》;20111231;第29卷;第79-103页 * |
| 基于分数阶微分的图像边缘细节检测与提取;陈青等;《电子学报》;20131031;第41卷(第10期);第1873-1880页 * |
| 基于多尺度边缘响应函数的自适应阈值边缘检测算法;岳思聪等;《电子与信息学报》;20080430;第30卷(第4期);第957-960页 * |
| 基于高斯小波的多尺度积图像边缘检测算法;赵晓丽等;《数据采集与处理》;20120731;第27卷(第4期);第490-494页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104156956A (en) | 2014-11-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107330437B (en) | Feature extraction method based on convolutional neural network target real-time detection model | |
| CN104156693B (en) | A kind of action identification method based on the fusion of multi-modal sequence | |
| CN105069807B (en) | A kind of stamped workpieces defect inspection method based on image procossing | |
| CN109427066A (en) | Edge detection method for any angle | |
| CN101981582B (en) | Method and apparatus for detecting object | |
| CN108121991B (en) | Deep learning ship target detection method based on edge candidate region extraction | |
| CN109902715B (en) | Infrared dim target detection method based on context aggregation network | |
| CN108898610A (en) | A kind of object contour extraction method based on mask-RCNN | |
| CN102819740B (en) | A kind of Single Infrared Image Frame Dim targets detection and localization method | |
| CN111640125A (en) | Mask R-CNN-based aerial photograph building detection and segmentation method and device | |
| CN106936964B (en) | A kind of mobile phone screen angular-point detection method based on Hough transformation template matching | |
| CN105740910A (en) | Vehicle object detection method and device | |
| CN107808161A (en) | A kind of Underwater targets recognition based on light vision | |
| CN113627257B (en) | Detection method, detection system, device and storage medium | |
| CN102324099A (en) | Step edge detection method oriented to humanoid robot | |
| CN102147867A (en) | Method for identifying traditional Chinese painting images and calligraphy images based on subject | |
| Li et al. | Research on a product quality monitoring method based on multi scale PP-YOLO | |
| Yang et al. | Semantic segmentation in architectural floor plans for detecting walls and doors | |
| CN104156956B (en) | A kind of multicorner edge detection operator method recognized based on Gauss wavelet one-dimensional peak value | |
| CN103971362B (en) | SAR image change-detection based on rectangular histogram and elite genetic algorithm for clustering | |
| CN110288026A (en) | A kind of image partition method and device practised based on metric relation graphics | |
| CN104463896A (en) | Image corner point detection method and system based on kernel similar region distribution characteristics | |
| Dulecha et al. | Crack detection in single-and multi-light images of painted surfaces using convolutional neural networks | |
| CN108597589A (en) | Model generating method, object detection method and medical image system | |
| Xu et al. | A deep neural network-based intelligent detection model for manufacturing defects of automobile parts |
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 |