CN113674174A - Line scanning cylinder geometric correction method and device based on significant row matching - Google Patents
Line scanning cylinder geometric correction method and device based on significant row matching Download PDFInfo
- Publication number
- CN113674174A CN113674174A CN202110965505.1A CN202110965505A CN113674174A CN 113674174 A CN113674174 A CN 113674174A CN 202110965505 A CN202110965505 A CN 202110965505A CN 113674174 A CN113674174 A CN 113674174A
- Authority
- CN
- China
- Prior art keywords
- line
- image
- matching
- corrected
- reference image
- 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
- 238000012937 correction Methods 0.000 title claims abstract description 48
- 238000000034 method Methods 0.000 title claims abstract description 23
- 238000005070 sampling Methods 0.000 claims abstract description 66
- 230000001131 transforming effect Effects 0.000 claims abstract description 7
- 230000004044 response Effects 0.000 claims description 27
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 5
- 238000003384 imaging method Methods 0.000 description 9
- 238000000605 extraction Methods 0.000 description 7
- 239000000463 material Substances 0.000 description 6
- 230000007547 defect Effects 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 4
- 230000006399 behavior Effects 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 239000003054 catalyst Substances 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000008602 contraction Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000013519 translation Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/80—Geometric correction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a line scanning cylinder geometric correction method and a device based on salient line matching, wherein the method comprises the following steps: step S1, using a line scan camera to perform line scan on a cylindrical workpiece to be corrected, acquiring an expanded image as an image to be corrected, and acquiring a reference image; step S2, carrying out gradient projection on the reference image and the image to be corrected; step S3, significant line selection is carried out according to the gradient projection value; step S4, for each significant line, searching a candidate matching line corresponding to the significant line on the image to be corrected; step S5, sampling the matching line pairs to obtain q matching line pairs, and determining the optimal sampling matching line pair through multiple iterations; step S6, with the optimal sampling matching line pair as reference, allocating a matching line on the image to be corrected for each salient line on the reference image; and step S7, transforming the image to be corrected by using the matching lines to obtain an image aligned with the reference image, thereby realizing geometric correction.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a method and a device for geometrically correcting a line-scanning cylinder based on significant row matching.
Background
In the industrial field, in order to detect surface defects of some cylindrical workpieces, it is necessary to image the surface thereof. Common applications include surface defect detection of lipsticks, cylindrical pencils, and the like.
The main methods currently used for imaging cylinders are: the workpiece is placed on a mechanism capable of enabling the workpiece to rotate, a line scanning camera is fixed above the cylindrical workpiece, the cylindrical workpiece is rotated at a constant speed through the mechanism, and signals are sent to the line scanning camera to acquire images. However, the cylindrical imaging acquired in this way presents two problems:
a. the starting position of the cylinder deployment differs each time.
The reason is as follows: the position of the face-to-face line scanning camera corresponds to the starting position of the unfolding when the workpiece is placed, and in fact, the cylindrical workpiece is difficult to be placed at the same position every time.
b. The cylindrical image is distorted in the direction of the unfolding.
The reason is as follows: since the rotation angular velocity of the mechanism is not absolutely uniform, the image obtained by the line scan camera is distorted. In this way the cylindrical unfolded image will be elongated or shortened to different degrees and the stretching is non-linear.
Based on these two problems, cylindrical workpieces are difficult to align using conventional image matching techniques as planar workpiece imaging, and thus large-scale inspection using machine vision is not possible.
In order to solve the above problems, there is also a technique of reducing image distortion by making the angular velocity more uniform by using an ultra-high precision mechanism device, but it is difficult to popularize at an excessively high cost.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide a geometric correction method and a geometric correction device for a line-scanned cylinder based on significant line matching, which achieve the aim of aligning a cylinder image obtained by line scanning to a reference image based on significant line matching, so that the same batch of cylinder materials can be aligned reasonably, and the cylinder materials can also be subjected to large-scale visual defect detection like plane materials.
To achieve the above and other objects, the present invention provides a method for correcting a geometric shape of a line-scan cylinder based on saliency matching, comprising the steps of:
step S1, acquiring an unfolded image of a perfect and undamaged cylindrical workpiece by using a line-scan camera, preprocessing the image to obtain an ideal image as a reference image, and performing line-scan acquisition on the unfolded image of the cylindrical workpiece to be corrected by using the line-scan camera as an image to be corrected;
step S2, respectively carrying out gradient projection on the reference image and the image to be corrected;
step S3, significant line selection is carried out on the reference image and the image to be corrected according to the gradient projection value;
step S4, for each significant line of the reference image, finding a candidate matching line corresponding to the significant line on the image to be corrected;
step S5, sampling the matching line pairs obtained in the step S4 to obtain q matching line pairs, correcting the vertical coordinate of each significant line on the image to be corrected according to the significant line reference of the reference image of each matching line pair, searching whether a significant line exists in the corresponding range on the reference image according to the correction result, updating the confidence coefficient of the sampling matching line pairs in the current group according to the search result, and determining the optimal sampling matching line pair through multiple iterations and according to the confidence coefficient;
step S6, with the optimal sampling matching line pair as reference, allocating a matching line on the image to be corrected for each salient line on the reference image;
and step S7, transforming the image to be corrected by using the obtained matching lines to obtain an image aligned with the reference image, thereby realizing geometric correction.
Preferably, in step S1, if the side of the cylinder changes along the Y direction of the image, the width and height of the reference image are the same and different from those of the image to be corrected; if the side surface of the cylinder changes along the X direction of the image, the height of the reference image is the same as that of the image to be corrected, and the width of the reference image is different from that of the image to be corrected.
Preferably, in step S1, the cylinder is expanded in the Y direction by default, and if the cylinder is expanded in the X direction, the expanded image is rotated by 90 degrees.
Preferably, in step S2, the reference image and the image to be corrected are respectively subjected to gradient calculation along the Y direction of the image, and horizontal projection is performed to obtain a column vector with a length equal to the image height.
Preferably, in step S3, the line with the gradient projection value greater than the predetermined threshold is used as the saliency line.
Preferably, in step S4, the ith salient line S in the reference image is determinedR(i) And the kth significant line S in the image to be correctedW(k) The match response M (i, k) is calculated as:
and selecting p significant lines with the minimum matching response as candidate matching lines according to the matching response values of the significant lines and the significant lines in the image to be corrected, wherein R (S)R(i) J) represents the i-th salient line S of the reference image RR(i) Gray value of upper j point, W (S)W(k) J denotes the kth significant line S of the image W to be correctedW(k) The gray value of the upper j point.
Preferably, the step S5 further includes the steps of:
step S500, randomly extracting q different salient lines from the salient lines in the reference image, and sorting the salient lines from small to large according to the vertical coordinate;
step S501, for each extracted significant line, extracting matching lines randomly from p candidate matching lines according to probability;
step S502, for q significant rows in the reference image, if the ordinate of the ith significant row is smaller than the ordinate of the jth significant row (i, j ∈ {1,2, …, q } and i ≠ j), then the matching row extracted from the image to be corrected must also satisfy the condition, otherwise, the current sampling is discarded and the step S500 is executed again;
step S503, according to q sampled matching line pairs, taking the sampling line of the reference image as a reference, correcting the ordinate of each significant line on the image to be corrected, searching whether significant lines exist in the corresponding range on the reference image according to the correction result, and updating the confidence coefficient of the sampling matching pair in the current group according to the search result;
and step S504, returning to the step S500, iterating to the set iteration times, and saving the matching line pair with the maximum confidence coefficient as the optimal sampling matching line pair.
Preferably, in step S501, the following operations are performed:
step 1: randomly generating 1 integer between 1 and p, wherein the integer corresponds to the selected candidate matching row;
step2, randomly generating 1 value with the value range of [0,1], and when the value is less than the probability of the candidate matching line, the current candidate matching line is a sampled matching line; otherwise, step1 is re-executed until a matching line is selected that meets the condition.
Preferably, step S503 further includes:
after each group of sampling matching pairs is obtained, the confidence coefficient of the matching pair is initially 0;
correcting the ordinate of each salient line on the image to be corrected according to the sampling behavior standard of the reference image;
and after the corrected vertical coordinate is obtained, searching whether a remarkable line exists in a corresponding range on the reference image, and if so, increasing the current confidence by 1.
Preferably, in step S6, the position of each significant line of the reference image is transformed to obtain the corresponding position of the image to be corrected, and the best corresponding line is found by performing the matching again near the corresponding position.
In order to achieve the above object, the present invention further provides a line-scan cylinder geometry correction apparatus based on saliency row matching, including:
the image acquisition and processing unit is used for acquiring an unfolded image of a perfect and undamaged cylindrical workpiece by using the line-scan camera, preprocessing the image to obtain an ideal image as a reference image, and performing line-scan acquisition and unfolding on the cylindrical workpiece to be corrected by using the line-scan camera to obtain the unfolded image as an image to be corrected;
the gradient projection unit is used for respectively carrying out gradient projection on the reference image and the image to be corrected;
the salient line selection unit is used for carrying out salient line selection on the reference image and the image to be corrected according to the gradient projection value;
the rough matching unit is used for searching a candidate matching line corresponding to each salient line of the reference image on the image to be corrected;
the sampling correction optimizing unit is used for sampling the matching line pairs obtained by the coarse matching unit to obtain q matching line pairs, correcting the vertical coordinate of each significant line on the image to be corrected according to the significant behavior reference of the reference image of each matching line pair, searching whether a significant line exists in the corresponding range on the reference image according to the correction result, updating the confidence coefficient of the sampling matching line pairs in the current group according to the search result, and determining the optimal sampling matching line pair through multiple iterations and according to the confidence coefficient;
the fine matching unit is used for taking the optimal sampling matching line pair as a reference and distributing a matching line on the image to be corrected for each salient line on the reference image;
and the geometric correction unit is used for transforming the image to be corrected by using the obtained matching lines to obtain an image aligned with the reference image, so that geometric correction is realized.
Compared with the prior art, the invention discloses a geometric correction method and a device of a line-scanning cylinder based on salient line matching, which obtains a reference image and an image to be corrected by using a line-scanning camera, then respectively carries out gradient projection on the reference image and the image to be corrected, carries out salient line selection on the reference image and the image to be corrected according to a gradient projection value, searches a candidate matching line corresponding to each salient line on the image to be corrected on each salient line of the reference image, samples the obtained matching line pairs to obtain q matching line pairs, corrects the longitudinal coordinate of each salient line on the image to be corrected according to the salient line reference of the reference image of each matching line pair, searches whether the salient line exists in a corresponding range on the reference image according to a correction result, and updates the confidence coefficient of the current group of sampled matching line pairs according to the search result, determining an optimal sampling matching line pair according to the confidence degree through multiple iterations, allocating a matching line on the image to be corrected for each significant line on the reference image by taking the optimal sampling matching line pair as a reference, and finally transforming the image to be corrected by using the obtained matching line pair to obtain an image aligned with the reference image, so that the purpose of aligning the cylindrical image obtained by line scanning to the reference image is realized based on the significant line matching, the same batch of cylindrical materials can be aligned reasonably, and the cylindrical materials can also be subjected to large-scale visual defect detection like plane materials.
Drawings
FIG. 1 is a flow chart illustrating the steps of a method for correcting the geometry of a line-scan cylinder based on saliency row matching according to the present invention;
FIG. 2 is a flow chart of line-scan cylinder geometry correction based on saliency row matching in an embodiment of the present invention;
fig. 3 is a schematic structural diagram of a line-scan cylinder geometry correction apparatus based on saliency row matching according to the present invention.
Detailed Description
Other advantages and capabilities of the present invention will be readily apparent to those skilled in the art from the present disclosure by describing the embodiments of the present invention with specific embodiments thereof in conjunction with the accompanying drawings. The invention is capable of other and different embodiments and its several details are capable of modification in various other respects, all without departing from the spirit and scope of the present invention.
Fig. 1 is a flowchart illustrating steps of a method for correcting a geometric correction of a line-scan cylinder based on saliency row matching according to an embodiment of the present invention, and fig. 2 is a flowchart illustrating geometric correction of a line-scan cylinder based on saliency row matching according to an embodiment of the present invention. As shown in fig. 1 and fig. 2, the method for correcting the geometric shape of a line-scan cylinder based on saliency row matching of the present invention includes the following steps:
and step S1, acquiring an unfolded image of a perfect cylinder workpiece by using the line-scan camera, preprocessing the image to obtain an ideal image as a reference image, and acquiring the unfolded image as an image to be corrected by using the line-scan camera to perform line scanning on the cylinder workpiece to be corrected.
In the invention, a line scan camera is used for performing line scan on a cylindrical workpiece to acquire an unfolded image, and the acquired image is an image to be corrected and marked as W.
Similarly, a linear scanning camera is used for acquiring an unfolded image of a perfect cylinder workpiece, and a final desired ideal image is obtained through operations of manual translation, cutting and the like and is used as a reference image and is recorded as R. The reference image contains exactly the unfolded content of the cylindrical workpiece, and the image to be corrected contains the repeated cylindrical side content. The two differ only in the direction of deployment, but the content of the imaging does not change. The invention aims to correct an image to be corrected by taking a reference image as a reference.
In the embodiment of the invention, if the side surface of the cylinder changes along the Y direction of the image, the width and the height of the reference image are the same and different from those of the image to be corrected; if the side surface of the cylinder changes along the X direction of the image, the height of the reference image is the same as that of the image to be corrected, and the width of the reference image is different from that of the image to be corrected.
For convenience of description, it is assumed in the present invention that the cylindrical side surface is expanded and then varies along the Y direction of the image, and the heights of the reference image and the image to be corrected are respectively denoted as hRAnd hWThe widths of the reference image and the image to be corrected are the same and are collectively denoted as w. The operation is similar when the cylinder side is unfolded and then changed in the X direction of the image.
Step S2, performing gradient projection on the reference image and the image to be corrected respectively.
Specifically, the reference image and the image to be corrected are respectively subjected to gradient calculation along the Y direction of the image, and horizontal projection is performed to obtain a column vector with the length being the image height.
Taking the reference image R as an example, the ith row gradient projection PR(i) Comprises the following steps:
the gradient projection calculation method for the image to be corrected is the same, and is not described herein again.
It should be noted that, in the present invention, the cylinder is developed in the Y direction by default, and if the cylinder is developed in the X direction, the developed image needs to be rotated by 90 degrees, so as to ensure that there is no difference between the reference image and the image content to be corrected in the X direction, and there is a difference in expansion and contraction only in the Y direction.
And step S3, performing salient line selection on the reference image and the image to be corrected according to the gradient projection value.
Since the line-by-line matching is not only computationally intensive, but also results in numerous mismatches due to the presence of a large number of insignificant lines. Therefore, the present invention screens out relatively significant rows. In the embodiment of the present invention, the threshold h is manually settThe gradient projection value is larger than a threshold value htThe row of (2) is taken as the saliency row.
When the significant line selection is executed on the reference image and the image to be corrected, a group of display line numbers are obtained and recorded as SRAnd SWThe number of elements is m and n. The ordinate value for the ith salient line in the reference image is denoted as yR(SR(i) Y) for the y-th significant line of the image to be correctedW(SW(j) ). The storage of the salient rows needs to be sorted from small to large by the ordinate.
Step S4, for each significant line of the reference image, finding a candidate matching line corresponding to the significant line on the image to be corrected.
The step of rough matching is to match each salient line of the reference image with each salient line of the image to be corrected, and to search p salient lines with the minimum matching response on the image to be corrected as candidate matching lines. Specifically, the ith salient line S in the reference image is divided intoR(i) With the kth significant line S in the image to be correctedW(k) The match response M (i, k) is calculated as:
R(SR(i) j) represents the i-th salient line S of the reference image RR(i) Gray value of upper j point, W (S)W(k) J denotes the kth significant line S of the image W to be correctedW(k) The gray value of the upper j point is,
and selecting p significant rows with the minimum matching response as candidate matching rows according to the matching response values of the significant rows in the reference image and the significant rows in the image to be corrected, wherein a parameter p is set by a user and is usually 2-4, normalizing (dividing by the total sum) the p matching responses to make the sum of the p matching responses be 1, and subtracting the p normalized values by 1 to obtain the matching probability of each candidate matching row. In the present invention, the 1 is used to ensure that the smaller the matching response, the higher the probability. Examples are as follows: assuming the ith significant line on the reference image, 2(p ═ 2) minimum matching response behaviors a and b on the image to be corrected, whose specific response values are M (i, a) ═ 100 and M (i, b) ═ 300, then after normalization, the two values are 100/(100+300) ═ 0.25,300/(100+300) ═ 0.75, so that the matching probability is 1-0.25 ═ 0.75 and 1-0.75 ═ 0.25, so that the matching probability of the matching response row a is 0.75 and the matching probability of b is 0.25, which means that the a row on the image to be corrected and the i row of the reference image match with a high probability and a probability of 0.75.
And step S5, sampling the matching line pairs obtained in the step S4 to obtain q matching line pairs, and correcting the ordinate of the corresponding significant line on the image to be corrected according to the significant line standard of the reference image of each matching line pair.
After the rough matching in step S4, each significant row of the reference image and the corresponding candidate matching row in the image to be corrected form a matching row pair, and the matching row pair obtained by the rough matching has different degrees of mismatching, but there are a large number of correct matching row pairs, so it is necessary to further determine the reliability of each matching row pair. In the embodiment of the invention, the matching line pair is sampled in a random sampling mode, and the reliability of the matching line pair is judged. The sampling for one time needs to be performed for multiple times until a proper matching row pair is acquired, and the step S5 specifically includes the following steps:
step S500, randomly extracting q different salient lines from the salient lines in the reference image as sampling lines, and sorting the sampling lines from small to large according to the ordinate. The parameter q is set by the user, and its value is related to the actually extracted saliency line and the imaging quality. When the number of the remarkable rows is large and the imaging quality is good, the number of the remarkable rows can be set to be larger, and the number of the remarkable rows needs to be smaller, and the value range is usually 5-30.
Step S501, for each extracted sampling line, randomly extracting matching lines from p candidate matching lines according to probability, wherein the specific operation is as follows:
step 1: randomly generating 1 integer with the value range of 1,2, …, p, wherein the value corresponds to the selected candidate matching row;
step2, randomly generating 1 value with the value range of [0,1], and when the value is less than the matching probability of the candidate matching row, the current candidate matching row is a sampled matching row; otherwise, step1 is re-executed until a matching line is selected that meets the condition.
Step S502, verifying the sequence consistency. For q significant lines (sampling lines) in the reference image, if the ordinate of the ith significant line is smaller than the ordinate of the jth significant line (i, j ∈ {1,2, …, q } and i ≠ j), the matching line extracted on the image to be corrected must also satisfy the condition, otherwise, the sampling is discarded and the step S500 is executed again.
After the sampling matching pair operation is performed, q matching row pairs are obtained.
Step S503, according to the q sampled matching line pairs, taking the sampling line of the reference image as a reference, correcting the ordinate of each significant line on the image to be corrected, searching whether significant lines exist in the corresponding range on the reference image according to the correction result, and updating the confidence of the current group of sampling matching pairs according to the search result.
Specifically, after each set of sample matching pairs (q matching row pairs are a set), the confidence of the initial set of sample matching pairs is 0. Suppose q matching line pairs obtained by sampling, where the ordinate of the q salient lines on the reference image is noted asThe ordinate of the matching line corresponding thereto on the image W to be corrected is recorded asThe ordinate of each salient line on the image to be corrected is corrected with reference to the sampling line (salient line) of the reference image (i.e. all salient lines are used to verify whether the currently selected set of matching line pairs is reasonable). With the jth significant row SW(j) For example, its original ordinate is yW(SW(j) Corrected ordinate is
After obtaining the corrected ordinateThen, corresponding range on the reference image And if the inner search is carried out, increasing the current confidence coefficient by 1, wherein rtThe parameters set for the user are generally 2-5 rtThe smaller the value, the more stringent the determination, rtThe larger the value, the more relaxed the decision.
Step S504 returns to step S500 to perform iteration to the set iteration number (for example, 100), and a group of matching line pairs with the highest confidence is saved as the optimal sampling matching line pair, that is, a best group of matching groups is selected from all the matching groups, and each group is q matching line pairs.
In step S6, a matching line is allocated to each significant line on the reference image on the image to be corrected, with the optimal sampling matching line pair as a reference.
Suppose that the optimal sampled matched row pair is obtained by iterative optimization, where the ordinate of the q optimal matched rows on the reference image is noted asThe ordinate of the best matching line corresponding to the image W to be corrected is recorded asAnd (4) carrying out matching again near each salient line position of the reference image to find the best corresponding line according to the corresponding position of the image to be corrected obtained by transformation. With the ith salient line S on the reference imageR(i) For example, its original ordinate is yR(SR(i) On the image to be corrected, the ordinate corresponding thereto is(note that, here, instead of operating on the original reference image, the coordinates on the reference image are transformed to the coordinates on the image to be corrected, so its subscript is set to the reference image R and its superscript is set to the image to be corrected W):
For salient lines S on the reference imageR(i) Obtaining the corresponding position of the image to be correctedAfter, in the rangeInterior re-match search and SR(i) Corresponding optimal matching row, let k beInteger in range, match response value of Using the line with the minimum response value as the salient line S of the reference imageR(i) In the corresponding line of the image to be corrected, it is marked as S 'for convenience of description'W(i) With ordinate yW(S′W(i))。
And step S7, transforming the image to be corrected by using the obtained matching lines to obtain an image aligned with the reference image, thereby realizing geometric correction.
In the present invention, the line position u' of the u-th line of the aligned image C on the image W to be corrected is:
a. when u is<yR(SR(1) In the case of a hot press machine),
the proportional relationship between the ordinate positions of the 1 st and 2 nd significant lines is used to estimate the line u' on the image W to be corrected corresponding to it:
wherein, u' -yW(S′W(1) Denotes the u 'th to y' th lines on the image WW(S′W(1) Distance of rows); u-yR(SR(1) Represents the u-th to y-th lines on the image RR(SR(1) Distance of rows); y isW(S′W(1))-yW(S′W(2) Y on the image WW(S′W(1) Row y to yW(S′W(2) Distance of rows, yR(SR(1))-yR(SR(2) Y on the image R)R(SR(1) Row y to yR(SR(2) The distance of the rows, the present invention scales the distance of the adjacent rows to obtain the above equation.
b. When y isR(SR(i))≤u<yR(SR(i +1)), i is 1,2, … q-1
The proportional relation between the vertical coordinate positions of the ith and (i +1) th remarkable rows is used for estimating the row u' corresponding to the ith remarkable row on the image W to be corrected:
accordingly, u' -yW(S′W(i) Denotes the u 'th to y' th lines on the image WW(S′W(i) Distance of rows); u-yR(SR(i) Represents the u-th to y-th lines on the image RR(SR(i) Distance of rows); y isW(S′W(i))-yW(S′W(i +1)) represents the y-th on the image WW(S′W(i) Row y to yW(S′W(i +1)) distance of rows, yR(SR(i))-yR(SR(i +1)) represents the y-th on the image RR(SR(i) Row y to yR(SR(i +1)) row distance.
c. When u ≧ yR(SR(q)) in the presence of a catalyst,
the line u' corresponding to the q-1 th and the q-th significant line ordinate positions on the image W to be corrected is presumed by means of the proportional relationship:
accordingly, u' -yW(S′W(q)) represents the u 'th to y' th lines on the image WW(S′W(q)) a distance of the rows; u-yR(SR(q)) represents the u-th to y-th lines on the image RR(SR(q)) a distance of the rows; y isW(S′W(q))-yW(S′W(q-1)) represents the y-th on the image WW(S′W(q)) line to the y-thW(S′W(q-1)) distance of rows, yR(SR(q))-yR(SR(q-1)) represents the y-th image on the image RR(SR(q)) line to the y-thR(SR(q-1)) distance of rows.
When the corresponding relationship is obtained, the line u' may be a small value, and the gray values of all data of the u-th line in the alignment image C need to be obtained through linear interpolation:
wherein the content of the first and second substances,on the presentation image WV is the abscissa with the value 1,2, …, w, w is the width of the reference image and the image to be corrected.
Fig. 3 is a schematic structural diagram of a line-scan cylinder geometry correction apparatus based on saliency row matching according to the present invention. As shown in fig. 3, the present invention provides a line-scan cylinder geometry correction apparatus based on saliency row matching, including:
the image acquisition and processing unit 201 is configured to acquire an unfolded image of an intact cylindrical workpiece by using the line-scan camera, perform preprocessing to obtain an ideal image as a reference image, and perform line-scan acquisition on the unfolded image of the cylindrical workpiece to be corrected by using the line-scan camera as an image to be corrected.
In the present invention, the image acquisition processing unit 201 performs line scanning on the cylindrical workpiece by using the line scan camera to acquire an expanded image, and the acquired image is an image to be corrected and is denoted as W.
Similarly, the image acquisition processing unit 201 acquires an unfolded image of a perfect cylinder workpiece by using the line scan camera, and obtains a final desired ideal image as a reference image, which is denoted as R, through operations such as manual translation and cropping. The reference image contains exactly the unfolded content of the cylindrical workpiece, and the image to be corrected contains the repeated cylindrical side content. The two differ only in the direction of deployment, but the content of the imaging does not change. The invention aims to correct an image to be corrected by taking a reference image as a reference.
In the embodiment of the invention, if the side surface of the cylinder changes along the Y direction of the image, the width and the height of the reference image are the same and different from those of the image to be corrected; if the side surface of the cylinder changes along the X direction of the image, the height of the reference image is the same as that of the image to be corrected, and the width of the reference image is different from that of the image to be corrected.
For convenience of description, it is assumed in the present invention that the cylindrical side surface is expanded and then varies along the Y direction of the image, and the heights of the reference image and the image to be corrected are respectively denoted as hRAnd hWThe widths of the reference image and the image to be corrected are the same and are collectively denoted as w. When the side of the cylinder is unfolded and then changed along the X direction of the image, the operation mode is similar
A gradient projection unit 202, configured to perform gradient projection on the reference image and the image to be corrected respectively.
Specifically, the gradient projection unit 202 calculates gradients of the reference image and the image to be corrected along the Y direction of the image, and performs horizontal projection to obtain a column vector with a high length.
Taking the reference image R as an example, the ith row gradient projection PR(i) Comprises the following steps:
the gradient projection calculation method for the image to be corrected is the same, and is not described herein again.
It should be noted that, in the present invention, the cylinder is developed in the Y direction by default, and if the cylinder is developed in the X direction, the developed image needs to be rotated by 90 degrees, so as to ensure that there is no difference between the reference image and the image content to be corrected in the X direction, and there is a difference in expansion and contraction only in the Y direction.
And the salient line selection unit 203 is used for performing salient line selection on the reference image and the image to be corrected according to the gradient projection value.
Because, the line-by-line matching is not only computationally intensiveAnd many mismatches can occur because there are a large number of insignificant rows. Therefore, the present invention screens out relatively significant rows. In the embodiment of the present invention, the threshold h is manually settThe gradient projection value is larger than a threshold value htThe row of (2) is taken as the saliency row.
When the significant line selection is executed on the reference image and the image to be corrected, a group of display line numbers are obtained and recorded as SRAnd SWThe number of elements is m and n. The ordinate value for the ith salient line in the reference image is denoted as yR(SR(i) Y) for the y-th significant line of the image to be correctedW(SW(j) ). The storage of the salient rows needs to be sorted from small to large by the ordinate.
And a rough matching unit 204, configured to find, for each significant line of the reference image, a candidate matching line corresponding to the significant line on the image to be corrected.
Specifically, the rough matching unit 204 matches each significant line of the reference image with each significant line of the image to be corrected, and finds p significant lines with the smallest matching response on the image to be corrected as candidate matching lines. Specifically, the ith salient line S in the reference image is divided intoR(i) With the kth significant line S in the image to be correctedW(k) The match response M (i, k) is calculated as:
and therefore, for each salient line of the reference image, selecting p salient lines with the minimum matching response as candidate matching lines according to the matching response values of the salient lines and the salient lines in the image to be corrected, wherein a parameter p is set by a user and is usually 2-4, normalizing (dividing by the total sum) the p matching responses to make the sum of the p matching responses be 1, and subtracting the p normalized values by 1 to obtain the matching probability of each candidate matching line. In the present invention, the 1 is used to ensure that the smaller the matching response, the higher the probability.
The sampling correction optimizing unit 205 is configured to sample the matching line pairs obtained by the coarse matching unit 204 to obtain q matching line pairs, correct the ordinate of each significant line on the image to be corrected according to the significant line reference of the reference image of each matching line pair, search whether a significant line exists in the corresponding range on the reference image according to the correction result, update the confidence of the current set of sampling matching line pairs according to the search result, and determine the optimal sampling matching line pair through multiple iterations and according to the confidence.
After rough matching by the rough matching unit 204, each significant line of the reference image and the corresponding candidate matching line in the image to be corrected form a matching line pair, and the matching line pair obtained by the rough matching has different degrees of mismatching, but a large number of correct matching line pairs also exist, so that the reliability of each matching line pair needs to be further determined. In the embodiment of the invention, the matching line pair is sampled in a random sampling mode, and the reliability of the matching line pair is judged. A sampling is performed for a plurality of times until a suitable matching row pair is obtained, and the sampling correction optimizing unit 205 includes:
and the reference image salient line extraction module is used for randomly extracting q different salient lines from the salient lines in the reference image and sorting the salient lines from small to large according to the ordinate. The parameter q is set by the user, and its value is related to the actually extracted saliency line and the imaging quality. When the number of the remarkable rows is large and the imaging quality is good, the number of the remarkable rows can be set to be larger, and the number of the remarkable rows needs to be smaller, and the value range is usually 5-30.
A candidate matching line extraction module, configured to extract, for each extracted significant line, a matching line from p candidate matching lines thereof according to probability, where the candidate matching line extraction module is specifically operative to:
step 1: randomly generating 1 integer with the value range of 1,2, …, p, wherein the value corresponds to the selected candidate matching row;
step2, randomly generating 1 value with the value range of [0,1], and when the value is less than the matching probability of the candidate matching row, the current candidate matching row is a sampled matching row; otherwise, step1 is re-executed until a matching line is selected that meets the condition.
And the sequence consistency verification module is used for performing sequence consistency verification on the extraction result of the candidate matching line extraction module, specifically, for q significant lines in the reference image, if the ordinate of the ith significant line is smaller than the ordinate of the jth significant line (i, j belongs to {1,2, …, q } and i ≠ j), the matching line extracted from the image to be corrected must also satisfy the condition, otherwise, the current sampling is discarded and returned to the reference image significant line extraction module.
After the sampling matching pair operation is performed, q matching row pairs are obtained.
And the salient line correction module is used for correcting the ordinate of each salient line on the image to be corrected by taking the sampling line of the reference image as a reference according to the q sampled matching line pairs, searching whether the salient line exists in the corresponding range on the reference image according to the correction result, and updating the confidence coefficient of the sampling matching pair in the current group according to the search result.
Specifically, after each set of sample matching pairs is obtained, the confidence of the initial set of sample matching pairs is 0. Suppose q matching line pairs obtained by sampling, where the ordinate of the q salient lines on the reference image is noted asThe ordinate of the matching line corresponding thereto on the image W to be corrected is recorded asThe ordinate of each salient line on the image to be corrected is corrected with reference to the sampling line (salient line) of the reference image. With the jth significant row SW(j) For example, its original ordinate is yW(SW(j) Corrected ordinate is
After obtaining the corrected ordinateThen, corresponding range on the reference image And if the inner search is carried out, increasing the current confidence coefficient by 1, wherein rtThe parameters set for the user are generally 2-5 rtThe smaller the value, the more stringent the determination, rtThe larger the value, the more relaxed the decision.
And the iteration execution module returns the reference image significant line extraction module to iterate to a set iteration number (for example, 100), and stores a group of matching line pairs with the highest confidence as an optimal sampling matching line pair, namely, selects a best matching group from all matching groups, wherein each group is q matching line pairs.
And the fine matching unit 206 is configured to allocate a matching line on the image to be corrected for each significant line on the reference image with reference to the optimal sampling matching line pair.
It is assumed that the sampling correction optimizing unit obtains an optimal sampling matching line pair by iterative optimization, wherein the ordinate of q optimal matching lines on the reference image is recorded asThe ordinate of the best matching line corresponding to the image W to be corrected is recorded asAnd (4) carrying out matching again near each salient line position of the reference image to find the best corresponding line according to the corresponding position of the image to be corrected obtained by transformation. With the ith salient line S on the reference imageR(i) For example, its original ordinate is yR(SR(i) On the image to be corrected, the ordinate corresponding thereto is
For salient lines S on the reference imageR(i) Obtaining the corresponding position of the image to be correctedAfter, in the rangeInterior re-match search and SR(i) Corresponding optimal matching row, let k beInteger in range, match response value of Using the line with the minimum response value as the salient line S of the reference imageR(i) In the corresponding line of the image to be corrected, it is marked as S 'for convenience of description'W(i) With ordinate yW(S′W(i))。
And a geometric correction unit 207, configured to transform the image to be corrected by using the obtained matching line to obtain an image aligned with the reference image, so as to implement geometric correction.
In the present invention, the line position u' of the u-th line of the aligned image C on the image W to be corrected is:
a. when u is<yR(SR(1) In the case of a hot press machine),
the proportional relationship between the ordinate positions of the 1 st and 2 nd significant lines is used to estimate the line u' on the image W to be corrected corresponding to it:
wherein, u' -yW(S′W(1) Denotes the u 'th to y' th lines on the image WW(S′W(1) Distance of rows); u-yR(SR(1) Represents the u-th to y-th lines on the image RR(SR(1) Distance of rows); y isW(S′W(1))-yW(S′W(2) Y on the image WW(S′W(1) Row y to yW(S′W(2) Distance of rows, yR(SR(1))-yR(SR(2) Y on the image R)R(SR(1) Row y to yR(SR(2) The distance of the rows, the present invention scales the distance of the adjacent rows to obtain the above equation.
b. When y isR(SR(i))≤u<yR(SR(i +1)), i is 1,2, … q-1
The proportional relation between the vertical coordinate positions of the ith and (i +1) th remarkable rows is used for estimating the row u' corresponding to the ith remarkable row on the image W to be corrected:
accordingly, u' -yW(S′W(i) Denotes the u 'th to y' th lines on the image WW(S′W(i) Distance of rows); u-yR(SR(i) Represents the u-th to y-th lines on the image RR(SR(i) Distance of rows); y isW(S′W(i))-yW(S′W(i +1)) represents the y-th on the image WW(S′W(i) Row y to yW(S′W(i +1)) distance of rows, yR(SR(i))-yR(SR(i +1)) represents the y-th on the image RR(SR(i) Row y to yR(SR(i +1)) row distance.
c. When u ≧ yR(SR(q)) in the presence of a catalyst,
the line u' corresponding to the q-1 th and the q-th significant line ordinate positions on the image W to be corrected is presumed by means of the proportional relationship:
wherein, u' -yW(S′W(q)) represents the u 'th to y' th lines on the image WW(S′W(q)) a distance of the rows; u-yR(SR(q)) represents the u-th to y-th lines on the image RR(SR(q)) a distance of the rows; y isW(S′W(q))-yW(S′W(q-1)) represents the y-th on the image WW(S′W(q)) line to the y-thW(S′W(q-1)) distance of rows, yR(SR(q))-yR(SR(q-1)) represents the y-th image on the image RR(SR(q)) line to the y-thR(SR(q-1)) distance of rows.
When the corresponding relationship is obtained, the line u' may be a small value, and the gray values of all data of the u-th line in the alignment image C need to be obtained through linear interpolation:
wherein v is the value of the abscissa of 1,2, …, w.
The foregoing embodiments are merely illustrative of the principles and utilities of the present invention and are not intended to limit the invention. Modifications and variations can be made to the above-described embodiments by those skilled in the art without departing from the spirit and scope of the present invention. Therefore, the scope of the invention should be determined from the following claims.
Claims (10)
1. A line-scan cylinder geometric correction method based on salient line matching comprises the following steps:
step S1, acquiring an unfolded image of a perfect and undamaged cylindrical workpiece by using a line-scan camera, preprocessing the image to obtain an ideal image as a reference image, and performing line-scan acquisition on the unfolded image of the cylindrical workpiece to be corrected by using the line-scan camera as an image to be corrected;
step S2, respectively carrying out gradient projection on the reference image and the image to be corrected;
step S3, significant line selection is carried out on the reference image and the image to be corrected according to the gradient projection value;
step S4, for each significant line of the reference image, finding a candidate matching line corresponding to the significant line on the image to be corrected;
step S5, sampling the matching line pairs obtained in the step S4 to obtain q matching line pairs, correcting the vertical coordinate of each significant line on the image to be corrected according to the significant line reference of the reference image of each matching line pair, searching whether a significant line exists in the corresponding range on the reference image according to the correction result, updating the confidence coefficient of the sampling matching line pairs in the current group according to the search result, and determining the optimal sampling matching line pair through multiple iterations and according to the confidence coefficient;
step S6, with the optimal sampling matching line pair as reference, allocating a matching line on the image to be corrected for each salient line on the reference image;
and step S7, transforming the image to be corrected by using the obtained matching lines to obtain an image aligned with the reference image, thereby realizing geometric correction.
2. The line-scan cylinder geometry correction method based on saliency row matching as claimed in claim 1, characterized by: in step S1, if the side of the cylinder changes along the Y direction of the image, the width and height of the reference image are the same and different from those of the image to be corrected; if the side surface of the cylinder changes along the X direction of the image, the height of the reference image is the same as that of the image to be corrected, and the width of the reference image is different from that of the image to be corrected.
3. The line-scan cylinder geometry correction method based on saliency row matching as claimed in claim 2, characterized by: in step S1, the cylinder is expanded in the Y direction by default, and if the cylinder is expanded in the X direction, the expanded image is rotated by 90 degrees.
4. The line-scan cylinder geometry correction method based on saliency row matching as claimed in claim 1, characterized by: in step S3, the line with the gradient projection value greater than the predetermined threshold is regarded as the significant line.
5. The line-scan cylinder geometry correction method based on saliency row matching as claimed in claim 1, characterized by: in step S4, the ith salient line S in the reference image is determinedR(i) And the kth significant line S in the image to be correctedW(k) The match response M (i, k) is calculated as:
and selecting p significant lines with the minimum matching response as candidate matching lines according to the matching response values of the significant lines and the significant lines in the image to be corrected, wherein R (S)R(i) J) represents the i-th salient line S of the reference image RR(i) Gray value of upper j point, W (S)W(k) J denotes the kth significant line S of the image W to be correctedW(k) The gray value of the upper j point.
6. The method for geometric correction of line-scan cylinder based on saliency row matching as claimed in claim 1, wherein step S5 further comprises the steps of:
s500, randomly extracting q different salient lines from the salient lines in the reference image, and sorting the salient lines from small to large according to the vertical coordinate;
step S501, for each extracted significant line, extracting matching lines randomly from p candidate matching lines according to probability;
step S502, for q significant rows in the reference image, if the ordinate of the ith significant row is smaller than the ordinate of the jth significant row (i, j ∈ {1,2, …, q } and i ≠ j), then the matching row extracted from the image to be corrected must also satisfy the condition, otherwise, the current sampling is discarded and the step S500 is executed again;
step S503, according to q sampled matching line pairs, taking the sampling line of the reference image as a reference, correcting the ordinate of each significant line on the image to be corrected, searching whether significant lines exist in the corresponding range on the reference image according to the correction result, and updating the confidence coefficient of the sampling matching pair in the current group according to the search result;
and step S504, returning to the step S500, iterating to the set iteration times, and saving the matching line pair with the maximum confidence coefficient as the optimal sampling matching line pair.
7. The method for geometry correction of line-scan cylinder based on saliency row matching as claimed in claim 6, wherein in step S501, the following operations are performed:
step 1: randomly generating 1 integer between 1 and p, wherein the integer corresponds to the selected candidate matching row;
step2, randomly generating 1 value with the value range of [0,1], and when the value is less than the probability of the selected candidate matching line, the current candidate matching line is a sampled matching line; otherwise, step1 is re-executed until a matching line is selected that meets the condition.
8. The method for line-scan cylinder geometric correction based on saliency row matching as claimed in claim 7, wherein step S503 further comprises:
after each group of sampling matching pairs is obtained, the confidence coefficient of the matching pair is initially 0;
correcting the ordinate of each salient line on the image to be corrected according to the sampling behavior standard of the reference image;
and after the corrected vertical coordinate is obtained, searching whether a remarkable line exists in a corresponding range on the reference image, and if so, increasing the current confidence by 1.
9. The method as claimed in claim 8, wherein in step S6, each salient line position of the reference image is transformed to obtain the corresponding position of the image to be corrected, and the image to be corrected is re-matched to find the best corresponding line.
10. A line-scan cylinder geometry correction apparatus based on saliency row matching, comprising:
the image acquisition and processing unit is used for acquiring an unfolded image of a perfect and undamaged cylindrical workpiece by using the line-scan camera, preprocessing the image to obtain an ideal image as a reference image, and performing line-scan acquisition and unfolding on the cylindrical workpiece to be corrected by using the line-scan camera to obtain the unfolded image as an image to be corrected;
the gradient projection unit is used for respectively carrying out gradient projection on the reference image and the image to be corrected;
the salient line selection unit is used for carrying out salient line selection on the reference image and the image to be corrected according to the gradient projection value;
the rough matching unit is used for searching a candidate matching line corresponding to each salient line of the reference image on the image to be corrected;
the sampling correction optimizing unit is used for sampling the matching line pairs obtained by the coarse matching unit to obtain q matching line pairs, correcting the vertical coordinate of each significant line on the image to be corrected according to the significant behavior reference of the reference image of each matching line pair, searching whether a significant line exists in the corresponding range on the reference image according to the correction result, updating the confidence coefficient of the sampling matching line pairs in the current group according to the search result, and determining the optimal sampling matching line pair through multiple iterations and according to the confidence coefficient;
the fine matching unit is used for taking the optimal sampling matching line pair as a reference and distributing a matching line on the image to be corrected for each salient line on the reference image;
and the geometric correction unit is used for transforming the image to be corrected by using the obtained matching lines to obtain an image aligned with the reference image, so that geometric correction is realized.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110965505.1A CN113674174B (en) | 2021-08-23 | 2021-08-23 | Line scanning cylinder geometric correction method and device based on significant line matching |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110965505.1A CN113674174B (en) | 2021-08-23 | 2021-08-23 | Line scanning cylinder geometric correction method and device based on significant line matching |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113674174A true CN113674174A (en) | 2021-11-19 |
CN113674174B CN113674174B (en) | 2023-10-20 |
Family
ID=78544877
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110965505.1A Active CN113674174B (en) | 2021-08-23 | 2021-08-23 | Line scanning cylinder geometric correction method and device based on significant line matching |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113674174B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114782276A (en) * | 2022-04-29 | 2022-07-22 | 电子科技大学 | Resistivity imaging dislocation correction method based on adaptive gradient projection |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102236798A (en) * | 2011-08-01 | 2011-11-09 | 清华大学 | Image matching method and device |
CN103136760A (en) * | 2013-03-26 | 2013-06-05 | 华北电力大学(保定) | Multi sensor image matching method based on fast and daisy |
CN103456022A (en) * | 2013-09-24 | 2013-12-18 | 中国科学院自动化研究所 | High-resolution remote sensing image feature matching method |
CN103489191A (en) * | 2013-09-24 | 2014-01-01 | 中国科学院自动化研究所 | Method for detecting changes of remarkable target of remote sensing image |
CN104077782A (en) * | 2014-07-11 | 2014-10-01 | 中国科学院自动化研究所 | Satellite-borne remote sense image matching method |
CN107918927A (en) * | 2017-11-30 | 2018-04-17 | 武汉理工大学 | A kind of matching strategy fusion and the fast image splicing method of low error |
-
2021
- 2021-08-23 CN CN202110965505.1A patent/CN113674174B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102236798A (en) * | 2011-08-01 | 2011-11-09 | 清华大学 | Image matching method and device |
CN103136760A (en) * | 2013-03-26 | 2013-06-05 | 华北电力大学(保定) | Multi sensor image matching method based on fast and daisy |
CN103456022A (en) * | 2013-09-24 | 2013-12-18 | 中国科学院自动化研究所 | High-resolution remote sensing image feature matching method |
CN103489191A (en) * | 2013-09-24 | 2014-01-01 | 中国科学院自动化研究所 | Method for detecting changes of remarkable target of remote sensing image |
CN104077782A (en) * | 2014-07-11 | 2014-10-01 | 中国科学院自动化研究所 | Satellite-borne remote sense image matching method |
CN107918927A (en) * | 2017-11-30 | 2018-04-17 | 武汉理工大学 | A kind of matching strategy fusion and the fast image splicing method of low error |
Non-Patent Citations (1)
Title |
---|
杨明东: "基于匹配策略融合的低误差快速图像拼接算法", 《计算机应用研究》, vol. 36, no. 4, pages 1222 - 1227 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114782276A (en) * | 2022-04-29 | 2022-07-22 | 电子科技大学 | Resistivity imaging dislocation correction method based on adaptive gradient projection |
CN114782276B (en) * | 2022-04-29 | 2023-04-11 | 电子科技大学 | Resistivity imaging dislocation correction method based on adaptive gradient projection |
Also Published As
Publication number | Publication date |
---|---|
CN113674174B (en) | 2023-10-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110942458B (en) | Temperature anomaly defect detection and positioning method and system | |
JP4125393B2 (en) | Image fragment splicing device | |
CN111429494B (en) | Biological vision-based point cloud high-precision automatic registration method | |
CN107240130B (en) | Remote sensing image registration method, device and system | |
CN108009986B (en) | Fragment splicing method and device based on edge information | |
CN107862319B (en) | Heterogeneous high-light optical image matching error eliminating method based on neighborhood voting | |
WO2014203687A1 (en) | Image processing method, image processing device, and image processing program | |
CN108763575B (en) | Image control point automatic selection method based on image control point database | |
WO2015035462A1 (en) | Point feature based 2d-3d registration | |
TWI291543B (en) | A method and a system for creating a reference image using unknown quality patterns | |
CN114897705A (en) | Unmanned aerial vehicle remote sensing image splicing method based on feature optimization | |
US20150186422A1 (en) | Image processing apparatus, image processing method, and image processing program | |
CN107845066B (en) | Urban remote sensing image splicing method and device based on piecewise affine transformation model | |
CN107945120B (en) | Sample block based rotation and scaling image restoration method | |
CN113674174A (en) | Line scanning cylinder geometric correction method and device based on significant row matching | |
US7113652B2 (en) | System and method for using normalized gray scale pattern find | |
CN113658080B (en) | Linear scanning cylinder geometric correction method and device based on characteristic point matching | |
CN117635421A (en) | Image stitching and fusion method and device | |
JPH0935062A (en) | Image processing method | |
JP6919758B2 (en) | Image collation device | |
US8655080B2 (en) | Method and apparatus for identifying combinations of matching regions in images | |
CN111402281B (en) | Book edge detection method and device | |
US7085433B2 (en) | System and method for performing rotational and translational testing for a reference image used in a normalized gray scale pattern find system | |
Balletti et al. | Image matching for historical maps comparison | |
Wang et al. | Automatic corresponding control points selection for historical document image registration |
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 |