CN107993193A - The tunnel-liner image split-joint method of surf algorithms is equalized and improved based on illumination - Google Patents
The tunnel-liner image split-joint method of surf algorithms is equalized and improved based on illumination Download PDFInfo
- Publication number
- CN107993193A CN107993193A CN201710857638.0A CN201710857638A CN107993193A CN 107993193 A CN107993193 A CN 107993193A CN 201710857638 A CN201710857638 A CN 201710857638A CN 107993193 A CN107993193 A CN 107993193A
- Authority
- CN
- China
- Prior art keywords
- mrow
- mtd
- image
- point
- illumination
- 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
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000005286 illumination Methods 0.000 title claims abstract description 49
- 238000009826 distribution Methods 0.000 claims abstract description 18
- 230000006978 adaptation Effects 0.000 claims abstract description 10
- 230000011218 segmentation Effects 0.000 claims abstract description 7
- 230000006835 compression Effects 0.000 claims description 10
- 238000007906 compression Methods 0.000 claims description 10
- 238000001914 filtration Methods 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 5
- 238000001514 detection method Methods 0.000 claims description 3
- 230000014509 gene expression Effects 0.000 claims description 3
- 239000000463 material Substances 0.000 claims description 3
- 238000013178 mathematical model Methods 0.000 claims description 3
- 230000003287 optical effect Effects 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000004321 preservation Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000000926 separation method Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims 1
- 230000000694 effects Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 230000000717 retained effect Effects 0.000 description 3
- 238000007689 inspection Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical group [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 229910002804 graphite Inorganic materials 0.000 description 1
- 239000010439 graphite Substances 0.000 description 1
- 230000005641 tunneling Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4038—Image mosaicing, e.g. composing plane images from plane sub-images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/10—Image enhancement or restoration using non-spatial domain filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- 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
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The tunnel-liner image split-joint method of surf algorithms is equalized and improved based on illumination, and this method includes:(1), fitting illumination curve distribution algorithm tunnel picture is handled, prominent features and the influence for reducing uneven illumination;(2), using sruf Feature Correspondence Algorithms find sruf features, and the above-mentioned characteristic point detected is matched using FlannBasedMatcher adaptations;(3), using symmetry matching strategy, distance matching limits algorithm, and to epipolar-line constraint the methods of carries out matching constraint;(4), ask for using mathematical way the desired distance of image, image is subjected to segmentation splicing by image.While the present invention can remove high light, retain edge of crack details.Separated tunnel each several part can be connected into an entirety, can accurately detect the length in longer crack, and a secondary more complete tunnel plate image is provided, facilitate the accurate positionin of various problems.
Description
Technical field:
The invention belongs to technical field of computer vision, more particularly to a kind of tunnel-liner merging algorithm for images.
Background technology:
Shielding force at direction of pump spindle whether balance influence to canned motor pump can safe and reliable operation.Axial force is by thrust disc with sliding
The axial end face friction pair of bearing undertakes, and axial force is excessive, causes bearing wear to aggravate, and axial gap becomes larger, and makes shielding pump shaft
The service life held rapidly reduces, even result in electric pump can not normal operation, cause the accident.Machine halt trouble occurs for canned motor pump, greatly
Partly cause is all due to that axial force is not balanced effectively or graphite bearing damages.It is thus accurate to shielding force at direction of pump spindle
Really, it is efficient to calculate, and be effectively balanced axial force, service life and safe and reliable fortune for canned motor pump
Row has great significance.
The content of the invention:
Goal of the invention:
The present invention provides a kind of tunnel-liner image split-joint method for being equalized and being improved surf algorithms based on illumination, its mesh
Be solve in the past the problems of.
Technical solution:
A kind of tunnel-liner image split-joint method for equalizing and improving surf algorithms based on illumination, it is characterised in that:Should
Method includes:
(1), the algorithm for being fitted illumination curve distribution is handled tunnel picture, and prominent features simultaneously reduce uneven illumination
Influence;
(2), sruf features are found using sruf Feature Correspondence Algorithms, and will using FlannBasedMatcher adaptations
The above-mentioned characteristic point detected is matched;
(3), using symmetry matching strategy, distance matching limits algorithm, and to epipolar-line constraint the methods of carries out matching constraint;
(4), the desired distance of image is asked for using mathematical way, image is subjected to segmentation splicing by image.
(1) in step, the algorithm steps for reducing the i.e. balanced uneven illumination of influence of uneven illumination are as follows:
Improved mean filter denoising, gaussian pyramid compression denoising and improved gaussian filtering fitting illumination curve point
Cloth.
(3) characteristic matching limitation uses following three kinds of schemes in step:
Symmetry matching strategy, distance matching limit algorithm and stochastical sampling consistency algorithm.
This method step is as follows:
(1) algorithm for being fitted illumination curve distribution is handled tunnel picture, and prominent features simultaneously reduce uneven illumination
Influence;
(1.1) improved mean filter denoising:
To remove spuious bright noise, mean filter is carried out to original image using (2*t+1) * (2*t+1) templates, by average
Filtered image takes the minimum value of each respective pixel to form new image compared with each grey scale pixel value of original image;
Retain FRACTURE CHARACTERISTICS while removing high bright;
G (i, j)=min [s (i, j), r (i, j)]
Wherein n=m=45;I, j are defined in the transverse and longitudinal coordinate on image, and s (i, j) is that original image coordinate is (i, j) place
Grey scale pixel value, r (i, j) is the grey scale pixel value that image coordinate after mean filter is (i, j) place, and g (i, j) is equal to improve
Image coordinate after value filtering processing is the grey scale pixel value at (i, j) place;
(1.2), gaussian pyramid compression denoising:
Compression of images is carried out using the image pyramid compress mode of some resolution ratio multi-scale expressions;Different levels correspond to
The image of different resolution;Image material particular is not lost as far as possible while noise is removed, and can improve computing speed
Degree, while the gradient in crack and background is retained, eliminates small echo moving noise;
Wherein, by the way that to input original image i stackings, for above step, obtained image is i+1 layers of image.w(m,
N) it is Gaussian convolution core, gl-1(i, j) is the grey scale pixel value that the i-th tomographic image coordinate is (i, j) place, gl(i, j) is i+1 layer
Image coordinate is the grey scale pixel value at (i, j) place.
(1.3), improved gaussian filtering fitting illumination curve distribution:
Illumination curve distribution is fitted, is smoothed equivalent to gray scale section curve figure, by section and one
Gaussian filter carries out convolution to realize;
X (i, j)=g (i, j) * h (i, j)
X (i, j) is to do convolution with Gaussian filter h (i, j) and image g (i, j) to estimate optical field distribution situation;
Make o (i, j)=t (i, j)-x (i, j)
Wherein i, j are defined in the transverse and longitudinal coordinate on image, and t (i, j) is that image coordinate is (i, j) place after pyramidal compression
Grey scale pixel value, x (i, j) is the grey scale pixel value that image coordinate is (i, j) place after gaussian filtering, and o (i, j) is removes illumination
Image coordinate is the grey scale pixel value at (i, j) place after curved surface, then compared with the image average after removing illumination curved surface, is retained
Less than the pixel value of equal value part, so as to filter out noise, y (i, j) is the pixel grey scale that image coordinate is (i, j) place after pre-processing
Value;
(2), sruf features are found using sruf Feature Correspondence Algorithms, and will using FlannBasedMatcher adaptations
The above-mentioned characteristic point detected is matched:
(2.1), SurfFeatureDetector property detectors are created characteristic point detection is carried out to image, and will detection
Characteristic point out is stored in characteristic point array keypoints_1, keypoints_2 respectively;
(2.2), structure SurfDescriptorExtractor profilers describe SURF features;
(2.3), the use of FlannBasedMatcher adaptations is the descriptor based on scalable nearest neighbor algorithm Flann
Adaptation is matched the above-mentioned characteristic point detected;
(3), three kinds of schemes of characteristic matching limitation:
(3.1), symmetry matching strategy:Remove unstable match point;
(3.2), distance matching limits algorithm:
The particular content of algorithm is the speed v and camera frame per second fps obtained by outside, then time downlink of the car in a frame
Into distance be exactly s=v* (1/fps), due to wanting vehicle approximation to remain a constant speed movement, then garage into distance s and same position
The location of pixels difference x put on the image is the relation mapped one by one, we set this coefficient of relationship as k, then x=ks, due to by mistake
The presence of difference, the neighborhood section (x ± δ) that we set an x is defined as threshold interval;
(3.3), stochastical sampling consistency algorithm (RANSAC):
RANSAC estimates the ginseng of mathematical model by iterative manner from one group of observation data set comprising " point not in the know "
Number;It is a kind of uncertain algorithm --- it has certain probability to draw a rational result;It is necessary in order to improve probability
Improve iterations;
(4), image is carried out segmentation splicing by the relative distance that image is asked for using mathematical way, and algorithm thinking is:Calculate
The slope of line between each match pointThe mode of the most match point, that is, slope of slope is chosen, these matchings
Point is exactly correct;
(3.1) in step, remove unstable match point algorithm realizes that process is as follows:
1) index of each match point in above-mentioned matching is found;
2) judge that the match point under the relative indexing in matching twice is opposite;
3) corresponding point in matching twice is preserved;
4) a each call number of match point is preserved.
(3.2) in step, which realizes that process is as follows:
1) with secondary each opposite match point of traversal, and the lateral separation between two match points is calculated;
2) limited using threshold value, the cycle through taking pictures, the travel speed of car also have the distance three of vehicle distances wall into
Row the Fitting Calculation goes out suitable distance threshold (x ± δ).
(3.3) in step, the basic assumption of RANSAC is:
1) data are made of " intra-office point ";
2) " point not in the know " is the data for not adapting to the model;
3) data in addition belong to noise;
Not in the know Producing reason has:The extreme value of noise;The measuring method of mistake;To the false supposition of data;
RANSAC reaches target by one group of random subset being chosen in data;The subset being selected is assumed to be
Intra-office point, and verified with following methods:
1) there is the intra-office point that a model is adapted to hypothesis, i.e., all unknown parameters can be calculated from the intra-office point of hypothesis
Draw;
2) gone to test all other data with the model obtained in 1, if some point is suitable for the model of estimation, it is believed that
It is also intra-office point;
3) if enough points are classified as the intra-office point of hypothesis, then the model of estimation is just reasonable enough;
4) then, go to reevaluate model with the intra-office point of all hypothesis, because it is only by initial hypothesis intra-office point
Estimated;
5) finally, by estimating the error rate of intra-office point and model come assessment models;
This process is repeatedly executed fixed number, otherwise the model produced every time is because intra-office point is given up very little
Abandon, otherwise it is selected because of more preferable than existing model;The algorithm realizes process:
1) characteristic point keypoints, is changed into the points types of coordinate form by transformation matrix;
1) basis matrix of image is calculated;
2) screening of exterior point in carrying out, point in preservation, eliminates exterior point.
(4) process of realizing of step is:
(4.1) find characteristic point above and matched;
(4.2) the distance between characteristic point among calculating two images;
(4.3) obtained distance is planned using the mode in mathematical algorithm;
(4.4) image among the second width image is translated;
(4.5) image after processing is spliced.
Advantageous effect:
The present invention provides a kind of tunnel-liner image split-joint method for being equalized and being improved surf algorithms based on illumination, this hair
It is bright while can remove high light, retain edge of crack details.Separated tunnel each several part can be connected into an entirety,
The length in longer crack can be accurately detected, and a secondary more complete tunnel plate image is provided, facilitates the accurate positionin of various problems.
Brief description of the drawings:
Fig. 1 a are artwork;
Fig. 1 b are section curve figure;
Fig. 1 c are pretreatment image;
Fig. 1 d scheme for matching;
Fig. 1 e are spliced map.
Embodiment:
The present invention relates to a kind of tunnel-liner image split-joint method for equalizing and improving surf algorithms based on illumination, such as scheme
Shown in 1, this method includes:It is proposed that a kind of algorithm for being fitted illumination curve distribution is handled tunnel picture, prominent features are simultaneously
Reduce the influence of uneven illumination;Sruf features are found using sruf Feature Correspondence Algorithms, and use FlannBasedMatcher
Orchestration is matched the above-mentioned characteristic point detected;Using symmetry matching strategy, distance matching limits algorithm, to polar curve about
The methods of beam, carries out matching constraint;The desired distance of image is asked for using mathematical way, image is subjected to segmentation splicing by image.
Comprise the following steps that:
1. shown in the original image figure (a) gathered in tunnel, wherein a certain row are chosen, is drawn and cutd open according to grey scale pixel value
Surface curve figure, such as schemes shown in (b), it is seen that not only have the Curvature Effect of illumination curve distribution, also containing substantial amounts of noise.Therefore examine
Worry first pre-processes image, proposes a kind of algorithm of balanced uneven illumination.Algorithm realizes process.
1.1 improved mean filter denoisings
The characteristics of what is presented in the picture due to crack is to have low gray value, and high gradient is poor.Therefore, higher than equal in image
The gray value of value must be noise, to remove this spuious bright noise, by the image after mean filter and each pixel of original image
Gray value is compared, and takes the minimum value of each respective pixel to form new image.Because the gray value in crack is relatively low, illumination ash
Angle value is higher, and institute while removing high bright so as to retain FRACTURE CHARACTERISTICS.
G (i, j)=min [s (i, j), r (i, j)]
Wherein n=m=45.I, j are defined in the transverse and longitudinal coordinate on image, and s (i, j) is that original image coordinate is (i, j) place
Grey scale pixel value, r (i, j) is the grey scale pixel value that image coordinate after mean filter is (i, j) place, and g (i, j) is equal to improve
Image coordinate after value filtering processing is the grey scale pixel value at (i, j) place.
1.2 gaussian pyramids compress denoising
Due to crack, to have width to have narrow, the gradient difference contrast between crack and background have by force have it is weak, therefore, using some points
The image pyramid compress mode of resolution multi-scale expression carries out compression of images.Different levels correspond to the image of different resolution.
This method does not lose image material particular as far as possible while noise is removed, and can improve arithmetic speed, is split in reservation
While seam and the gradient of background, small echo moving noise is eliminated.
1.3 improved gaussian filtering fitting illumination curve distributions
Illumination curve distribution is fitted, is smoothed equivalent to gray scale section curve figure, this can be by section
Convolution is carried out with a Gaussian filter to realize.Illumination patterns uniformization effect after the algorithm process is good.Pre- place
Image after reason is as schemed shown in (c).
X (i, j)=g (i, j) * h (i, j)
X (i, j) is to do convolution with Gaussian filter h (i, j) and image g (i, j) to estimate optical field distribution situation.
Make o (i, j)=t (i, j)-x (i, j)
Wherein i, j are defined in the transverse and longitudinal coordinate on image, and t (i, j) is that image coordinate is (i, j) place after pyramidal compression
Grey scale pixel value, x (i, j) is the grey scale pixel value that image coordinate is (i, j) place after gaussian filtering, and o (i, j) is removes illumination
Image coordinate is the grey scale pixel value at (i, j) place after curved surface, then compared with the image average after removing illumination curved surface, is retained
Less than the pixel value of equal value part, so as to filter out noise, y (i, j) is the pixel grey scale that image coordinate is (i, j) place after pre-processing
Value.
2. sruf features are found using sruf Feature Correspondence Algorithms, and will be upper using FlannBasedMatcher adaptations
The characteristic point detected is stated to be matched
2.1 create SurfFeatureDetector property detectors, and (SURFFeatureDetector is in OpenCV2
Generic features detector introduces a kind of new general-purpose interface and is used for different detectors.It this interface define a KeyPoint
Class is to encapsulate the attribute of each characteristic point.For Harris angle points, only position is useful), characteristic point inspection is carried out to image
Survey, and the characteristic point detected is stored in characteristic point array keypoints_1, keypoints_2 respectively.
2.2 structure SurfDescriptorExtractor profilers (are equally the generic features inspections in OpenCV 2
Survey device and introduce a kind of new general-purpose interface, carry out feature description), SURF features are described.
2.3 use the FlannBasedMatcher adaptation (descriptors match based on scalable nearest neighbor algorithm Flann
Device) the above-mentioned characteristic point detected is matched.
3. three kinds of schemes of characteristic matching limitation
3.1 symmetry matching strategies
The principle of above-mentioned Feature Points Matching is gone and the characteristic point on another piece image with the characteristic point on piece image
Matching, finds the characteristic point of matching probability maximum as match point, we guess, the second width figure of matching is removed by piece image
Obtained match point should with by the second width figure go the obtained match point of the first width figure of matching should be it is consistent, it is still, actual
On twice matched result be not consistent, so according to this consistent principle above, those unstable can be removed
With point, which realizes process:
5) index of each match point in above-mentioned matching is found.
6) judge that the match point under the relative indexing in matching twice is opposite.
7) corresponding point in matching twice is preserved.
8) a each call number of match point is preserved.
3.2 distance matchings limit algorithm
The tunneling features point that narrow sense is changed to by the limitation of extensive Feature Points Matching matches limitation, and then we think, according to car
Contact between speed and the frame per second taken pictures, estimates a shift length on image, is limited afterwards into row distance, it should
Matched error rate can be reduced.The particular content of algorithm is the speed v and camera frame per second fps obtained by outside, then car exists
The distance advanced under the time of one frame is exactly s=v* (1/fps), due to wanting vehicle approximation to remain a constant speed movement, then garage into
Distance s and the location of pixels difference x of same position on the image be the relation mapped one by one, we set this coefficient of relationship as k,
Then x=ks, due to the presence of error, the neighborhood section (x ± δ) that we set an x is defined as threshold interval.The calculation
Method realizes process:
1) with secondary each opposite match point of traversal, and the lateral separation between two match points is calculated.
2) limited using threshold value, the cycle through taking pictures, the travel speed of car also have the distance three of vehicle distances wall into
Row the Fitting Calculation goes out suitable distance threshold (x ± δ).
3.3 stochastical sampling consistency algorithms (RANSAC)
RANSAC can estimate mathematical model from one group of observation data set comprising " point not in the know " by iterative manner
Parameter.It is a kind of uncertain algorithm --- it has certain probability to draw a rational result;Must in order to improve probability
Iterations must be improved.
The basic assumption of RANSAC is:
2) data are made of " intra-office point ", such as:The distribution of data can be explained with some model parameters;
2) " point not in the know " is the data for not adapting to the model;
3) data in addition belong to noise.
Not in the know Producing reason has:The extreme value of noise;The measuring method of mistake;To the false supposition of data.
RANSAC reaches target by one group of random subset being chosen in data.The subset being selected is assumed to be
Intra-office point, and verified with following methods:
1) there is the intra-office point that a model is adapted to hypothesis, i.e., all unknown parameters can be calculated from the intra-office point of hypothesis
Draw.
2) gone to test all other data with the model obtained in 1, if some point is suitable for the model of estimation, it is believed that
It is also intra-office point.
3) if enough points are classified as the intra-office point of hypothesis, then the model of estimation is just reasonable enough.
4) then, go to reevaluate model with the intra-office point of all hypothesis, because it is only by initial hypothesis intra-office point
Estimated.
5) finally, by estimating the error rate of intra-office point and model come assessment models.
This process is repeatedly executed fixed number, otherwise the model produced every time is because intra-office point is given up very little
Abandon, otherwise it is selected because of more preferable than existing model.The algorithm realizes process:
1) characteristic point keypoints, is changed into the points types of coordinate form by transformation matrix.
3) basis matrix of image is calculated.
4) screening of exterior point in carrying out, point in preservation, eliminates exterior point.
Image result after matching is as schemed (d).
4. asking for the relative distance of image using mathematical way, image is subjected to segmentation splicing by image, algorithm thinking is:
Since the probability that error hiding occurs in matching process is very high, but for substantial amounts of matched data, this point point is very
It is small, in order to increase accuracy, reduce the influence of mistake, we calculate the slope of line between each match point here(note:Opposite characteristic point is respectively (x on two width figuresi,yi) and (xj,yj)) due to correct match point herein it
Between line it is always parallel, and mistake match point between slope it is widely different since error connection line is seldom, we select
The most match point of the slope i.e. mode of slope is taken, these match points are exactly correct.The process of realization is:
4.1 find characteristic point above and are matched.
The distance between characteristic point among 4.2 calculating two images.
4.3 are planned obtained distance using the mode in mathematical algorithm
4.4 are translated the image among the second width image
4.5 are spliced the image after processing
Spliced image effect is as schemed (e).
Claims (8)
- A kind of 1. tunnel-liner image split-joint method for equalizing and improving surf algorithms based on illumination, it is characterised in that:The party Method includes:(1), the algorithm for being fitted illumination curve distribution is handled tunnel picture, prominent features and the shadow for reducing uneven illumination Ring;(2), sruf features are found using sruf Feature Correspondence Algorithms, and will be above-mentioned using FlannBasedMatcher adaptations The characteristic point detected is matched;(3), using symmetry matching strategy, distance matching limits algorithm, and to epipolar-line constraint the methods of carries out matching constraint;(4), the desired distance of image is asked for using mathematical way, image is subjected to segmentation splicing by image.
- 2. the tunnel-liner image split-joint method according to claim 1 for equalizing and improving surf algorithms based on illumination, It is characterized in that:(1) in step, the algorithm steps for reducing the i.e. balanced uneven illumination of influence of uneven illumination are as follows:Improved mean filter denoising, gaussian pyramid compression denoising and improved gaussian filtering fitting illumination curve distribution.
- 3. the tunnel-liner image split-joint method according to claim 1 for equalizing and improving surf algorithms based on illumination, It is characterized in that:(3) characteristic matching limitation uses following three kinds of schemes in step:Symmetry matching strategy, distance matching limit algorithm and stochastical sampling consistency algorithm.
- 4. the tunnel-liner image split-joint method according to claim 1 for equalizing and improving surf algorithms based on illumination, It is characterized in that:This method step is as follows:(1) algorithm for being fitted illumination curve distribution is handled tunnel picture, prominent features and the influence for reducing uneven illumination;(1.1) improved mean filter denoising:To remove spuious bright noise, mean filter is carried out to original image using (2*t+1) * (2*t+1) templates, by mean filter Image afterwards takes the minimum value of each respective pixel to form new image compared with each grey scale pixel value of original image;Going Retain FRACTURE CHARACTERISTICS while except high light;<mrow> <mi>r</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfrac> <mn>1</mn> <mrow> <mo>(</mo> <mn>2</mn> <mi>n</mi> <mo>+</mo> <mn>1</mn> <mo>)</mo> <mo>*</mo> <mo>(</mo> <mn>2</mn> <mi>m</mi> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> </mfrac> <munderover> <mo>&Sigma;</mo> <mrow> <mi>t</mi> <mo>=</mo> <mo>-</mo> <mi>n</mi> </mrow> <mi>n</mi> </munderover> <munderover> <mo>&Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mo>-</mo> <mi>m</mi> </mrow> <mi>m</mi> </munderover> <mi>s</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>-</mo> <mi>t</mi> <mo>,</mo> <mi>j</mi> <mo>-</mo> <mi>k</mi> <mo>)</mo> </mrow> </mrow>G (i, j)=min [s (i, j), r (i, j)]Wherein n=m=45;I, j are defined in the transverse and longitudinal coordinate on image, and s (i, j) is the picture that original image coordinate is (i, j) place Plain gray value, r (i, j) are that the image coordinate after mean filter is the grey scale pixel value at (i, j) place, and g (i, j) is improvement average filter Image coordinate after ripple processing is the grey scale pixel value at (i, j) place;(1.2), gaussian pyramid compression denoising:Compression of images is carried out using the image pyramid compress mode of some resolution ratio multi-scale expressions;Different levels correspond to different The image of resolution ratio;Image material particular is not lost as far as possible while noise is removed, and can improve arithmetic speed, While the gradient of reservation crack and background, small echo moving noise is eliminated;<mrow> <mi>w</mi> <mo>=</mo> <mfrac> <mn>1</mn> <mn>256</mn> </mfrac> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <mn>1</mn> </mtd> <mtd> <mn>4</mn> </mtd> <mtd> <mn>6</mn> </mtd> <mtd> <mn>4</mn> </mtd> <mtd> <mn>1</mn> </mtd> </mtr> <mtr> <mtd> <mn>4</mn> </mtd> <mtd> <mn>16</mn> </mtd> <mtd> <mn>24</mn> </mtd> <mtd> <mn>16</mn> </mtd> <mtd> <mn>4</mn> </mtd> </mtr> <mtr> <mtd> <mn>6</mn> </mtd> <mtd> <mn>24</mn> </mtd> <mtd> <mn>36</mn> </mtd> <mtd> <mn>24</mn> </mtd> <mtd> <mn>6</mn> </mtd> </mtr> <mtr> <mtd> <mn>4</mn> </mtd> <mtd> <mn>16</mn> </mtd> <mtd> <mn>24</mn> </mtd> <mtd> <mn>16</mn> </mtd> <mtd> <mn>4</mn> </mtd> </mtr> <mtr> <mtd> <mn>1</mn> </mtd> <mtd> <mn>4</mn> </mtd> <mtd> <mn>6</mn> </mtd> <mtd> <mn>4</mn> </mtd> <mtd> <mn>1</mn> </mtd> </mtr> </mtable> </mfenced> </mrow><mrow> <msub> <mi>g</mi> <mi>l</mi> </msub> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>m</mi> <mn>1</mn> <mo>=</mo> <mo>-</mo> <mn>2</mn> </mrow> <mn>2</mn> </munderover> <munderover> <mo>&Sigma;</mo> <mrow> <mi>n</mi> <mn>1</mn> <mo>=</mo> <mo>-</mo> <mn>2</mn> </mrow> <mn>2</mn> </munderover> <mi>w</mi> <mrow> <mo>(</mo> <mi>m</mi> <mn>1</mn> <mo>,</mo> <mi>n</mi> <mn>1</mn> <mo>)</mo> </mrow> <munderover> <mo>&Sigma;</mo> <mrow> <mi>m</mi> <mo>=</mo> <mo>-</mo> <mn>2</mn> </mrow> <mn>2</mn> </munderover> <munderover> <mo>&Sigma;</mo> <mrow> <mi>n</mi> <mo>=</mo> <mo>-</mo> <mn>2</mn> </mrow> <mn>2</mn> </munderover> <mi>w</mi> <mrow> <mo>(</mo> <mi>m</mi> <mo>,</mo> <mi>n</mi> <mo>)</mo> </mrow> <msub> <mi>q</mi> <mrow> <mi>l</mi> <mo>-</mo> <mn>2</mn> </mrow> </msub> <mrow> <mo>(</mo> <mn>2</mn> <mo>&times;</mo> <mo>(</mo> <mrow> <mn>2</mn> <mi>i</mi> <mo>+</mo> <mi>m</mi> <mn>1</mn> </mrow> <mo>)</mo> <mo>+</mo> <mi>m</mi> <mo>,</mo> <mn>2</mn> <mo>&times;</mo> <mo>(</mo> <mrow> <mn>2</mn> <mi>j</mi> <mo>+</mo> <mi>n</mi> <mn>1</mn> </mrow> <mo>)</mo> <mo>+</mo> <mi>n</mi> <mo>)</mo> </mrow> </mrow><mrow> <msub> <mi>g</mi> <mi>l</mi> </msub> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>m</mi> <mo>=</mo> <mo>-</mo> <mn>2</mn> </mrow> <mn>2</mn> </munderover> <munderover> <mo>&Sigma;</mo> <mrow> <mi>n</mi> <mo>=</mo> <mo>-</mo> <mn>2</mn> </mrow> <mn>2</mn> </munderover> <mi>w</mi> <mrow> <mo>(</mo> <mi>m</mi> <mo>,</mo> <mi>n</mi> <mo>)</mo> </mrow> <msub> <mi>g</mi> <mrow> <mi>l</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> <mrow> <mo>(</mo> <mn>2</mn> <mi>i</mi> <mo>+</mo> <mi>m</mi> <mo>,</mo> <mn>2</mn> <mi>j</mi> <mo>+</mo> <mi>n</mi> <mo>)</mo> </mrow> </mrow>Wherein, by the way that to input original image i stackings, for above step, obtained image is i+1 layers of image;W (m, n) is Gaussian convolution core, gl-1(i, j) is the grey scale pixel value that the i-th tomographic image coordinate is (i, j) place, gl(i, j) is i+1 tomographic image Coordinate is the grey scale pixel value at (i, j) place;(1.3), improved gaussian filtering fitting illumination curve distribution:Illumination curve distribution is fitted, is smoothed equivalent to gray scale section curve figure, by section and a Gauss Smoothing filter carries out convolution to realize;<mrow> <mi>h</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfrac> <mn>1</mn> <mrow> <mn>2</mn> <msup> <mi>&pi;&sigma;</mi> <mn>2</mn> </msup> </mrow> </mfrac> <msup> <mi>e</mi> <mfrac> <mrow> <mo>-</mo> <mrow> <mo>(</mo> <msup> <mi>i</mi> <mn>2</mn> </msup> <mo>+</mo> <msup> <mi>j</mi> <mn>2</mn> </msup> <mo>)</mo> </mrow> </mrow> <mrow> <mn>2</mn> <msup> <mi>&sigma;</mi> <mn>2</mn> </msup> </mrow> </mfrac> </msup> </mrow>X (i, j)=g (i, j) * h (i, j)X (i, j) is to do convolution with Gaussian filter h (i, j) and image g (i, j) to estimate optical field distribution situation;Make o (i, j)=t (i, j)-x (i, j)<mrow> <mi>y</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>=</mo> <mi>min</mi> <mo>&lsqb;</mo> <mfrac> <mn>1</mn> <mrow> <mo>(</mo> <mn>2</mn> <mi>n</mi> <mo>+</mo> <mn>1</mn> <mo>)</mo> <mo>*</mo> <mo>(</mo> <mn>2</mn> <mi>m</mi> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> </mfrac> <munderover> <mo>&Sigma;</mo> <mrow> <mi>t</mi> <mo>=</mo> <mo>-</mo> <mi>n</mi> </mrow> <mi>n</mi> </munderover> <munderover> <mo>&Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mo>-</mo> <mi>m</mi> </mrow> <mi>m</mi> </munderover> <mi>o</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>-</mo> <mi>t</mi> <mo>,</mo> <mi>j</mi> <mo>-</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>,</mo> <mi>o</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>&rsqb;</mo> </mrow>Wherein i, j are defined in the transverse and longitudinal coordinate on image, and t (i, j) is the picture that image coordinate is (i, j) place after pyramidal compression Plain gray value, x (i, j) are the grey scale pixel value that image coordinate is (i, j) place after gaussian filtering, and o (i, j) is removes illumination curved surface Image coordinate is the grey scale pixel value at (i, j) place afterwards, then compared with the image average after removing illumination curved surface, reservation is less than The pixel value of equal value part, so as to filter out noise, y (i, j) is the grey scale pixel value that image coordinate is (i, j) place after pre-processing;(2), sruf features are found using sruf Feature Correspondence Algorithms, and will be above-mentioned using FlannBasedMatcher adaptations The characteristic point detected is matched:(2.1), SurfFeatureDetector property detectors are created and characteristic point detection is carried out to image, and will detected Characteristic point be stored in respectively in characteristic point array keypoints_1, keypoints_2;(2.2), structure SurfDescriptorExtractor profilers describe SURF features;(2.3), the use of FlannBasedMatcher adaptations is the descriptors match based on scalable nearest neighbor algorithm Flann Device is matched the above-mentioned characteristic point detected;(3), three kinds of schemes of characteristic matching limitation:(3.1), symmetry matching strategy:Remove unstable match point;(3.2), distance matching limits algorithm:The particular content of algorithm is the speed v and camera frame per second fps obtained by outside, then car is advanced under the time of a frame Distance is exactly s=v* (1/fps), due to want vehicle approximation to remain a constant speed movement, then garage into distance s and same position exist Location of pixels difference x on image is the relation mapped one by one, we set this coefficient of relationship as k, then x=ks, due to error In the presence of the neighborhood section (x ± δ) that we set an x is defined as threshold interval;(3.3), stochastical sampling consistency algorithm (RANSAC):RANSAC estimates the parameter of mathematical model by iterative manner from one group of observation data set comprising " point not in the know ";It It is a kind of uncertain algorithm --- it has certain probability to draw a rational result;It must be improved repeatedly to improve probability Generation number;(4), image is carried out segmentation splicing by the relative distance that image is asked for using mathematical way, and algorithm thinking is:Calculate each The slope of line between bar match pointThe mode of the most match point, that is, slope of slope is chosen, these match points are just It is correct;
- 5. the tunnel-liner image split-joint method according to claim 4 for equalizing and improving surf algorithms based on illumination, It is characterized in that:(3.1) in step, remove unstable match point algorithm realizes that process is as follows:1) index of each match point in above-mentioned matching is found;2) judge that the match point under the relative indexing in matching twice is opposite;3) corresponding point in matching twice is preserved;4) a each call number of match point is preserved.
- 6. the tunnel-liner image split-joint method according to claim 4 for equalizing and improving surf algorithms based on illumination, It is characterized in that:(3.2) in step, which realizes that process is as follows:1) with secondary each opposite match point of traversal, and the lateral separation between two match points is calculated;2) limited using threshold value, the cycle through taking pictures, the distance three that the travel speed of car also has vehicle distances wall is intended It is total to calculate suitable distance threshold (x ± δ).
- 7. the tunnel-liner image split-joint method according to claim 4 for equalizing and improving surf algorithms based on illumination, It is characterized in that:(3.3) in step, the basic assumption of RANSAC is:1) data are made of " intra-office point ";2) " point not in the know " is the data for not adapting to the model;3) data in addition belong to noise;Not in the know Producing reason has:The extreme value of noise;The measuring method of mistake;To the false supposition of data;RANSAC reaches target by one group of random subset being chosen in data;The subset being selected is assumed to be intra-office Point, and verified with following methods:1) there is the intra-office point that a model is adapted to hypothesis, i.e., all unknown parameters can be calculated from the intra-office point of hypothesis Go out;2) gone to test all other data with the model obtained in 1, if some point is suitable for the model of estimation, it is believed that it It is intra-office point;3) if enough points are classified as the intra-office point of hypothesis, then the model of estimation is just reasonable enough;4) then, go to reevaluate model with the intra-office point of all hypothesis, because it is only by initial hypothesis intra-office point estimation Cross;5) finally, by estimating the error rate of intra-office point and model come assessment models;This process is repeatedly executed fixed number, otherwise the model produced every time is wanted because intra-office point is rejected very little It is selected because of more preferable than existing model;The algorithm realizes process:1) characteristic point keypoints, is changed into the points types of coordinate form by transformation matrix;1) basis matrix of image is calculated;2) screening of exterior point in carrying out, point in preservation, eliminates exterior point.
- 8. the tunnel-liner image split-joint method according to claim 4 for equalizing and improving surf algorithms based on illumination, It is characterized in that:(4) process of realizing of step is:(4.1) find characteristic point above and matched;(4.2) the distance between characteristic point among calculating two images;(4.3) obtained distance is planned using the mode in mathematical algorithm;(4.4) image among the second width image is translated;(4.5) image after processing is spliced.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710857638.0A CN107993193B (en) | 2017-09-21 | 2017-09-21 | Tunnel lining image splicing method based on illumination equalization and surf algorithm improvement |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710857638.0A CN107993193B (en) | 2017-09-21 | 2017-09-21 | Tunnel lining image splicing method based on illumination equalization and surf algorithm improvement |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107993193A true CN107993193A (en) | 2018-05-04 |
CN107993193B CN107993193B (en) | 2021-06-11 |
Family
ID=62029758
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710857638.0A Active CN107993193B (en) | 2017-09-21 | 2017-09-21 | Tunnel lining image splicing method based on illumination equalization and surf algorithm improvement |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107993193B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110781731A (en) * | 2019-09-17 | 2020-02-11 | 上海电机学院 | Inspection robot instrument identification method based on specular reflection removal |
WO2021017699A1 (en) * | 2019-07-30 | 2021-02-04 | 广州市百果园信息技术有限公司 | Method for detecting abnormal point pairs, image splicing method, apparatus, and device |
CN117975374A (en) * | 2024-03-29 | 2024-05-03 | 山东天意机械股份有限公司 | Intelligent visual monitoring method for double-skin wall automatic production line |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101916371A (en) * | 2010-09-01 | 2010-12-15 | 北京工业大学 | Method for illuminating/normalizing image and method for identifying image by using same |
WO2013059966A1 (en) * | 2011-10-28 | 2013-05-02 | Telefonaktiebolaget L M Ericsson (Publ) | Protection in ring network of label switching routers |
CN104680516A (en) * | 2015-01-08 | 2015-06-03 | 南京邮电大学 | Acquisition method for high-quality feature matching set of images |
CN106934795A (en) * | 2017-01-23 | 2017-07-07 | 陕西师范大学 | The automatic testing method and Forecasting Methodology of a kind of glue into concrete beam cracks |
-
2017
- 2017-09-21 CN CN201710857638.0A patent/CN107993193B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101916371A (en) * | 2010-09-01 | 2010-12-15 | 北京工业大学 | Method for illuminating/normalizing image and method for identifying image by using same |
WO2013059966A1 (en) * | 2011-10-28 | 2013-05-02 | Telefonaktiebolaget L M Ericsson (Publ) | Protection in ring network of label switching routers |
CN104680516A (en) * | 2015-01-08 | 2015-06-03 | 南京邮电大学 | Acquisition method for high-quality feature matching set of images |
CN106934795A (en) * | 2017-01-23 | 2017-07-07 | 陕西师范大学 | The automatic testing method and Forecasting Methodology of a kind of glue into concrete beam cracks |
Non-Patent Citations (1)
Title |
---|
李海洋等: ""一种改进的基于SURF特征匹配的图像拼接算法", 《太原理工大学学报》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021017699A1 (en) * | 2019-07-30 | 2021-02-04 | 广州市百果园信息技术有限公司 | Method for detecting abnormal point pairs, image splicing method, apparatus, and device |
CN110781731A (en) * | 2019-09-17 | 2020-02-11 | 上海电机学院 | Inspection robot instrument identification method based on specular reflection removal |
CN117975374A (en) * | 2024-03-29 | 2024-05-03 | 山东天意机械股份有限公司 | Intelligent visual monitoring method for double-skin wall automatic production line |
Also Published As
Publication number | Publication date |
---|---|
CN107993193B (en) | 2021-06-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Rudolph et al. | Same same but differnet: Semi-supervised defect detection with normalizing flows | |
CN109190752B (en) | Image semantic segmentation method based on global features and local features of deep learning | |
CN110189255B (en) | Face detection method based on two-stage detection | |
CN103279765B (en) | Steel wire rope surface damage detection method based on images match | |
Li et al. | Single image depth estimation by dilated deep residual convolutional neural network and soft-weight-sum inference | |
CN105913415A (en) | Image sub-pixel edge extraction method having extensive adaptability | |
CN103440644A (en) | Multi-scale image weak edge detection method based on minimum description length | |
CN101140216A (en) | Gas-liquid two-phase flow type recognition method based on digital graphic processing technique | |
CN107993193A (en) | The tunnel-liner image split-joint method of surf algorithms is equalized and improved based on illumination | |
Lee et al. | Edge detection analysis | |
CN111738931B (en) | Shadow removal algorithm for aerial image of photovoltaic array unmanned aerial vehicle | |
Zhao et al. | Research on detection method for the leakage of underwater pipeline by YOLOv3 | |
CN110276747B (en) | Insulator fault detection and fault rating method based on image analysis | |
CN114972332A (en) | Bamboo laminated wood crack detection method based on image super-resolution reconstruction network | |
CN111160100A (en) | Lightweight depth model aerial photography vehicle detection method based on sample generation | |
CN113436216B (en) | Electrical equipment infrared image edge detection method based on Canny operator | |
Shit et al. | An encoder‐decoder based CNN architecture using end to end dehaze and detection network for proper image visualization and detection | |
CN107748885B (en) | Method for recognizing fuzzy character | |
CN117152076A (en) | Road surface crack detection method based on U-shaped full convolution neural network | |
CN117351078A (en) | Target size and 6D gesture estimation method based on shape priori | |
Wang et al. | Fast blur detection algorithm for UAV crack image sets | |
CN116452826A (en) | Coal gangue contour estimation method based on machine vision under shielding condition | |
Cheng et al. | Two-branch convolutional sparse representation for stereo matching | |
CN113834447B (en) | High-dynamic laser light bar self-adaptive imaging processing method in outdoor complex environment | |
CN113807206B (en) | SAR image target identification method based on denoising task assistance |
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 |