CN109711457A - It is a kind of based on improve the HU not rapid image matching method of bending moment and its application - Google Patents
It is a kind of based on improve the HU not rapid image matching method of bending moment and its application Download PDFInfo
- Publication number
- CN109711457A CN109711457A CN201811577536.4A CN201811577536A CN109711457A CN 109711457 A CN109711457 A CN 109711457A CN 201811577536 A CN201811577536 A CN 201811577536A CN 109711457 A CN109711457 A CN 109711457A
- Authority
- CN
- China
- Prior art keywords
- image
- matching
- template
- matched
- resolution ratio
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a kind of based on the HU not rapid image matching method of bending moment and its application is improved, and belongs to technical field of image processing.By utilizing pyramid model, the proportional resolution ratio for reducing template image and image to be matched, optimization starting matching position and searching route, reduces template and traverse calculation amount；It is proposed a kind of method for constructing multiple dimensioned template, and finally realize that target is accurately positioned and marks using measuring similarity and connected domain algorithm, so that the algorithm has certain invariance to grey scale change, translation, significantly rotation and scaling, and to big image in different resolution, matching speed greatly improves, and achieves good matching effect；More stable and robust is matched in tonal distortion, image focus, target translation, rotation and scaling, simultaneously, because the present invention overcomes the big disadvantages of big image in different resolution overall situation traversal matching primitives amount, matching efficiency is greatly improved, high-accuracy machining and the requirement of detection industry real-time matching have been substantially met.
Description
Technical field
The present invention relates to a kind of based on the HU not rapid image matching method of bending moment and its application is improved, and belongs to image procossing
Technical field.
Background technique
Images match is the hot issue in image procossing and computer vision field, is widely used in numerous areas, such as
Target identification, image retrieval, Text region, recognition of face, medical image analysis, robot navigation etc..Image matching method is big
Cause can be divided into two classes: first is that the matching based on frequency domain, such as Fourier transformation and wavelet transformation；Second is that being based on spatial domain
Matching, mainly include matching algorithm and feature-based matching algorithm based on region.
As some desirable features extract the appearance of operator, feature-based matching method is more widely applied, to image
The factors such as translation, rotation, scaling, grey scale change, which occur, has preferable robustness, becomes the emphasis of Recent study.
With going deep into for application, to the high-resolution target work in the fields such as high-accuracy machining and semiconductor detection
When part matches, traditional Feature Correspondence Algorithm has that computationally intensive, calculating cycle is longer, can not real-time matching；And
There are when multiple jamming targets, be unable to reach accurate matched requirement in a sub-picture.
Summary of the invention
In order to solve presently, there are the above problem, the present invention provides a kind of based on the rapid image for improving HU not bending moment
Matching process and its application.
The first purpose of this invention be to provide it is a kind of based on the rapid image matching method for improving HU not bending moment, it is described
Method the following steps are included:
S1, input template image T and image to be matched S；
S2 carries out image preprocessing to template image T and image to be matched S respectively；
S3, it is down-sampled to the progress of pretreated two images using image pyramid algorithm, it obtains that target spy can be recognized
The two images T ' and S ' of the lowest resolution of sign, and it is big according to the different scale of template image T using image pyramid algorithm
The small template for generating different scale size；
S4 carries out matching degree calculating using improved images match search strategy；
S5 searches optimal match point using similarity measurement method, realizes final goal positioning and label.
It is optionally, described down-sampled to the progress of pretreated two images using image pyramid algorithm, comprising:
Corresponding image is established using high-resolution template image T and image to be matched S as pyramidal bottom
Pyramid；The resolution ratio of the tomographic image second from the bottom of each image pyramid is the resolution ratio of bottom layer imageN-th layer reciprocal
The resolution ratio of image is the resolution ratio of bottom layer imageThe number of plies of each image pyramid is higher, image is smaller, resolution ratio
It is lower；The top of each image pyramid is the image that can recognize the lowest resolution of target signature, is denoted as image T ' respectively
With S '.
Optionally, the improved images match search strategy includes:
Step 1: calculating the mass center (x of image T ' and S '1,y1) and (x2,y2)；
Step 2: with the mass center (x to image S '2,y2) it is used as initial search point, image T ', which is overlayed image S ' above, to be made
(x1,y1) and (x2,y2) be overlapped；
Step 3: with (x2,y2) centered on, creation search subgraph (S, a, b) (- R≤a in the neighborhood of (p*p, radius R)
≤ R ,-R≤b≤R), and the similarity measure values of each pixel of matching primitives pixel-by-pixel in neighborhood, it is denoted as dab；Wherein, p is
The width and height of neighborhood；dabIndicate the matching degree of image T ' and S '；dabBe worth it is smaller, show image T ' and search subgraph (S, a, b)
Matching degree is higher；
Step 4: by dabAnd dabCorresponding pixel (a, b) is stored in a two-dimensional array, is denoted as B [r, c] (1≤r
≤ p, 1≤c≤p)；
Step 5: repeating step 3 and step 4, until all traversal terminates in neighborhood, find out the smallest similarity measurements of step 4
Magnitude dmin, by dminMass center coordinate on image S ' of the corresponding pixel (a, b) as best match position image.
Optionally, described image, which pre-processes, includes:
Step1: gray processing is carried out to image, converts grayscale image for cromogram；
Step2: gaussian filtering is carried out to grayscale image, the noise spot as caused by environment is removed, it is complete to obtain Edge preservation
The high image of contrast；
Step3: the image binaryzation that Step2 is obtained separates target with background, increases the specific gravity of useful information, drop
Low interference information specific gravity；
Step4: the image after the binaryzation obtained to Step3 carries out connected domain analysis processing.
Optionally, the similarity measure values calculation formula of the similarity measurement method is formula (1)
Wherein, Hu1 (k) (k=1~7) indicates the 7 invariant moments of template image, and Hu2 (k) (k=1~7) indicates search
The 7 invariant moments of figure.
Second object of the present invention is to provide a kind of above-mentioned based on the rapid image matching method for improving HU not bending moment
Application method in the processing detection of high-accuracy mechanical workpieces.
Optionally, the high-accuracy mechanical workpieces include: semiconductor chip, PCB circuit board, gear rotor, touch it is aobvious
Display screen, sensor.
Optionally, which comprises
The image of high-accuracy mechanical workpieces is obtained as image to be matched S using industrial camera shooting；
Selected from actual production process no defect, feature be obvious, workpiece product shooting picture of size qualification as
Template image T；
High-accuracy mechanical workpieces are detected based on the rapid image matching method for improving HU not bending moment using above-mentioned.
Optionally, the method also includes: before being detected, using template image T to high-accuracy mechanical workpieces into
Row positioning and tracking.
Optionally, the method also includes: after being detected, to high-accuracy mechanical workpieces carry out size detection and/
Or defect recognition.
The medicine have the advantages that
By utilizing pyramid model, the proportional resolution ratio for reducing template image and image to be matched proposes one kind
Improved matching search strategy, optimization starting matching position and searching route, reduce template and traverse calculation amount；Not for traditional HU
The low problem of bending moment matching algorithm precision, which proposes a kind of method for constructing multiple dimensioned template, and utilizes similarity degree
Amount and connected domain algorithm finally realize that target is accurately positioned and marks.Test result show the algorithm to grey scale change, translation,
Significantly rotating and scaling has certain invariance, and to big image in different resolution, matching speed is greatly improved, and achieves good
Matching effect；More stable and robust is matched in tonal distortion, image focus, target translation, rotation and scaling, especially
When being that significantly rotation and dimensional variation occur for target, possess more outstanding matching effect, matching precision reaches Pixel-level.
Meanwhile big image in different resolution overall situation traversal matching primitives amount is big to be lacked because image matching method proposed by the present invention overcomes
Point, greatly improves matching efficiency, has substantially met high-accuracy machining and the requirement of detection industry real-time matching.
Detailed description of the invention
To describe the technical solutions in the embodiments of the present invention more clearly, make required in being described below to embodiment
Attached drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for
For those of ordinary skill in the art, without creative efforts, it can also be obtained according to these attached drawings other
Attached drawing.
Fig. 1 is semiconductor chip experimental image.
Fig. 2 is gear stator experimental image.
Fig. 3 is big resolution ratio pcb board image.
Fig. 4 is improved search strategy schematic diagram.
Fig. 5 is improved HU not bending moment Fast Match Algorithm flow chart.
Fig. 6 is improved HU not testing result of the bending moment Fast Match Algorithm to pcb board.
Fig. 7 is improved HU not testing result of the bending moment Fast Match Algorithm to gear stator.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with attached drawing to embodiment party of the present invention
Formula is described in further detail.
Embodiment one:
The present embodiment provides a kind of based on the rapid image matching method for improving HU not bending moment, which comprises
S1, input template image T and image to be matched S；
S2 carries out image preprocessing to template image T and image to be matched S respectively；
S3, it is down-sampled to the progress of pretreated two images using image pyramid algorithm, it obtains that target spy can be recognized
The two images T ' and S ' of the lowest resolution of sign, and be the minimum resolution that can recognize target signature using image pyramid algorithm
The each image of rate generates the template of different scale size；
S4 carries out matching degree calculating using improved images match search strategy；
S5 searches optimal match point using similarity measurement method, realizes final goal positioning and label.
Wherein, down-sampled using image pyramid algorithm in step S3, comprising: by high-resolution template image T and
Image to be matched S establishes corresponding image pyramid respectively as pyramidal bottom；Each image pyramid it is second from the bottom
The resolution ratio of tomographic image is the resolution ratio of bottom layer imageThe resolution ratio of n-th layer image reciprocal is the resolution ratio of bottom layer imageThe number of plies of each image pyramid is higher, image is smaller, resolution ratio is lower；The top of each image pyramid is distinguishable
The image for knowing the lowest resolution of target signature, is denoted as image T ' and S ' respectively.
Improved images match search strategy in step S4 includes:
Step 1: calculating the mass center (x of image T ' and S '1,y1) and (x2,y2)；
Step 2: with the mass center (x to image S '2,y2) it is used as initial search point, image T ', which is overlayed image S ' above, to be made
(x1,y1) and (x2,y2) be overlapped；
Step 3: with (x2,y2) centered on, creation search subgraph (S, a, b) (- R≤a in the neighborhood of (p*p, radius R)
≤ R ,-R≤b≤R), and the similarity measure values of each pixel of matching primitives pixel-by-pixel in neighborhood, it is denoted as dab；Wherein, p is
The width and height of neighborhood；dabIndicate the matching degree of image T ' and S '；dabBe worth it is smaller, show image T ' and search subgraph (S, a, b)
Matching degree is higher；
The similarity measure values calculation formula of similarity measurement method is formula (1)
Wherein, Hu1 (k) (k=1~7) indicates the 7 invariant moments of template image, and Hu2 (k) (k=1~7) indicates search
The 7 invariant moments of figure.
Step 4: by dabAnd dabCorresponding pixel (a, b) is stored in a two-dimensional array, is denoted as B [r, c] (1≤r
≤ p, 1≤c≤p)；
Step 5: repeating step 3 and step 4, until all traversal terminates in neighborhood, find out the smallest similarity measurements of step 4
Magnitude dmin, by dminMass center coordinate on image S ' of the corresponding pixel (a, b) as best match position image.
Image preprocessing in step S2 includes:
Step1: gray processing is carried out to image, converts grayscale image for cromogram；
Step2: gaussian filtering is carried out to grayscale image, the noise spot as caused by environment is removed, it is complete to obtain Edge preservation
The high image of contrast；
Step3: the image binaryzation that Step2 is obtained separates target with background, increases the specific gravity of useful information, drop
Low interference information specific gravity；
Step4: the image after the binaryzation obtained to Step3 carries out connected domain analysis processing.
It is provided in this embodiment based on the rapid image matching method for improving HU not bending moment, by utilizing pyramid model,
The proportional resolution ratio for reducing template image and image to be matched, optimization starting matching position and searching route, reduce template
Traverse calculation amount；Method by constructing multiple dimensioned template, and target is finally realized using measuring similarity and connected domain algorithm
It is accurately positioned and marks, there is certain invariance to grey scale change, translation, significantly rotation and scaling, and to big resolution ratio
Image, matching speed greatly improve, and achieve good matching effect.
Embodiment two
Added based on the rapid image matching method for improving HU not bending moment in high-accuracy mechanical workpieces the present embodiment provides a kind of
Application method in work detection, the high-accuracy mechanical workpieces include: semiconductor chip, PCB circuit board, gear rotor, touching
Display screen, sensor etc. are touched, the present embodiment is illustrated by taking semiconductor chip, gear stator and pcb board as an example:
Because the intensity profile of image to be matched is uneven, the spy of big angle rotary, translation, scaling occurs for multiple target, target
Point makes the matching location difficulty of target object.And since the resolution ratio of image to be matched is larger, the matching primitives of target object
Amount is big, and matching positioning time is long.This example chooses that gray scale is uneven, gray scale changes, target object is out of focus partly leads respectively
Body chip image, target occur translation, rotation, the gear sub-image of scaling and the pcb board image with big resolution ratio and carry out
Object matching experiment.As shown in Figure 1, Figure 2, Figure 3 shows.
Image in the present embodiment is shot using industrial camera, provided by the invention constant based on HU is improved in application
Before the rapid image matching method of square detects high-accuracy mechanical workpieces, need to select not having from actual production process
Defect, feature be obvious, size qualification workpiece product shooting picture is as template image T；It is obtained using industrial camera high-accuracy
The image of mechanical workpieces is as image to be matched S；
The first step, input template image T and image to be matched S.
Second step carries out image preprocessing to two images T and S.
The image preprocessing, specifically: gray processing being carried out to image first, converts grayscale image for cromogram；Its
Secondary carry out gaussian filtering removes the noise spot as caused by environment, obtains the high image of the complete contrast of Edge preservation；Then will
Image binaryzation separates target with background, increases the specific gravity of useful information, reduces interference information specific gravity.Finally to binaryzation
Image afterwards carries out connected domain analysis, and after processing, image target area is more clear, and profile information is more obvious.
Third step, it is down-sampled using image pyramid algorithm, obtain two width that can recognize the lowest resolution of target signature
Image T ' and S ', and utilize the template of image pyramid 4 kinds of different scale sizes of generation.
The image pyramid algorithm is down-sampled, specifically: taking pyramidal bottom as the high-resolution of image to be processed
Rate indicates that tomographic image second from the bottom is the 1/4 of former resolution ratio, and tomographic image third from the bottom is the 1/16 of former resolution ratio, and so on,
The resolution ratio of n-th layer image reciprocal is the resolution ratio of bottom layer imageThe number of plies of each image pyramid is higher, image more
It is small, resolution ratio is lower；The top of each image pyramid is the image that can recognize the lowest resolution of target signature, is remembered respectively
Make image T ' and S '.
The template of the different scale size, specifically: according to the multi-scale expression characteristic of pyramid algorith, providing one kind
Multiple dimensioned template verifies matching precision of the HU invariant moments matching algorithm when graphical rule changes.HU square is carried out every time
When matching, according to image to be matched dimensional variation size, the template of 4 kinds of different scales is generated using image pyramid, then distinguish
It is matched with image to be matched.
4th step carries out matching degree calculating using improved images match search strategy.
Traditional images matching search strategy is all with the upper left corner of image to be matched (0,0) point as starting point, then entirely
Office traverses pixel-by-pixel.When starting point apart from best match position farther out, pixel-by-pixel in non-possible matching position meter when moving die plate
The match time that evaluation time occupies totality is excessive, and efficiency of algorithm is caused to be lower.
And the invention proposes a kind of improved images match search strategy method, schematic diagram is as shown in Figure 4.This method
The specific implementation steps are as follows:
(1) mass center (x of image T ' and S ' are calculated1,y1) and (x2,y2)；
(2) with the mass center (x to image S '2,y2) it is used as initial search point, image T ', which is overlayed image S ', above makes (x1,
y1) and (x2,y2) be overlapped；(3) with (x2,y2) centered on, creation search subgraph (S, a, b) in the neighborhood of (p*p, radius R)
(- R≤a≤R ,-R≤b≤R), and the similarity measure values of each pixel of matching primitives pixel-by-pixel in neighborhood, are denoted as dab；Its
In, p is the width and height of neighborhood；dabIndicate the matching degree of image T ' and S '；dabBe worth it is smaller, show image T ' and search subgraph
(S, a, b) matching degree is higher；
(4) by dabPixel (a, b) corresponding with the matching value is stored in a two-dimensional array, be denoted as B [r, c] (1≤
R≤p, 1≤c≤p)；
(5) step 3 and step 4 are repeated, terminates until all being traversed in neighborhood, finds out the smallest d in the 4th stepmin, corresponding
(a, b) be best match position image coordinate of the mass center on image S '.
The value of the R can determine according to specific circumstances.
The dminSolution, specifically: it is when beginning stepping through pixel in neighborhood, every row in B [r, c] is calculated
The minimum taking-up of matching value is stored in minimum sequence BminIn [r]；Finally the matching value in minimum value sequence is arranged
Sequence finds out the minimum value d in minimumminWith corresponding (a, b).Due to image S ' be it is known, search out the mass center of target
(a, b) coordinate on S ' it is known that coordinate of the mass center on image T ' it is known that in summary information can be accurately positioned target and exist
Position on S '.
5th step searches optimal match point using similarity measurement method, realizes final goal positioning and label.
The similarity measurement method, specifically:
In formula, Hu1 (k) (k=1~7) indicates the 7 invariant moments of template image；Hu2 (k) (k=1~7) indicates search
The 7 invariant moments of figure；dabIndicate matching value, dabBe worth it is smaller, show template image T and search subgraph (S, a, b) matching more connect
Closely.
As shown in figure 5, being algorithm flow chart, according to the process, image to be matched and template image is inputted, can be obtained
Match positioning result.
Fig. 6 gives the matching result of matching process provided by the invention, it will be appreciated from fig. 6 that match party provided by the invention
Method has very strong adaptability, and the figure of big angle rotary, translation, scaling occurs for, multiple target uneven for intensity profile, target
The matching of picture has preferable result.
Table 1 gives big resolution pcb board images match runing time result.It is provided according to the data present invention in table 1
Matching process match time it is short, speed is fast, high-efficient.
Table 1: big pcb board images match runing time result of differentiating is compared before and after algorithm improvement
The present embodiment for HU bending moment algorithm does not carry out images match when, the problems such as poor robustness is computationally intensive, and speed is slow,
Propose a kind of rapid image matching algorithm of improvement HU not bending moment.Experiments have shown that flat in gray scale unevenness, image focus, target
Move, rotation and scaling match that this paper matching algorithm is more stable and robust, especially target occur significantly rotation and
When dimensional variation, possess more outstanding matching effect.Meanwhile it being asked for what traditional HU invariant moments matching algorithm took a long time
Topic, the algorithm that the application proposes overcome the big disadvantage of big image in different resolution overall situation traversal matching primitives amount, greatly improve
With efficiency, high-accuracy machining and the requirement of detection industry real-time matching have been substantially met.
Part steps in the embodiment of the present invention, can use software realization, and corresponding software program can store can
In the storage medium of reading, such as CD or hard disk.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and
Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (10)
1. a kind of based on the rapid image matching method for improving HU not bending moment, which is characterized in that the described method comprises the following steps:
S1, input template image T and image to be matched S；
S2 carries out image preprocessing to template image T and image to be matched S respectively；
S3, it is down-sampled to the progress of pretreated two images using image pyramid algorithm, it obtains that target signature can be recognized
The two images T ' and S ' of lowest resolution, and using image pyramid algorithm according to the big your pupil of different scale of template image T
At the template of different scale size；
S4 carries out matching degree calculating using improved images match search strategy；
S5 searches optimal match point using similarity measurement method, realizes final goal positioning and label.
2. the method according to claim 1, wherein described utilize image pyramid algorithm to pretreated two
Width image carries out down-sampled, comprising:
Corresponding image gold word is established using high-resolution template image T and image to be matched S as pyramidal bottom
Tower；The resolution ratio of the tomographic image second from the bottom of each image pyramid is the resolution ratio of bottom layer imageN-th layer image reciprocal
Resolution ratio be bottom layer image resolution ratioThe number of plies of each image pyramid is higher, image is smaller, resolution ratio is lower；
The top of each image pyramid is the image that can recognize the lowest resolution of target signature, is denoted as image T ' and S ' respectively.
3. method according to claim 1 or 2, which is characterized in that the improved images match search strategy includes:
Step 1: calculating the mass center (x of image T ' and S '1,y1) and (x2,y2)；
Step 2: with the mass center (x to image S '2,y2) it is used as initial search point, image T ', which is overlayed image S ', above makes (x1,y1)
With (x2,y2) be overlapped；
Step 3: with (x2,y2) centered on, (p*p, radius R) neighborhood in creation search subgraph (S, a, b) (- R≤a≤
R ,-R≤b≤R), and the similarity measure values of each pixel of matching primitives pixel-by-pixel in neighborhood, it is denoted as dab；Wherein, p is neighbour
The width and height in domain；dabIndicate the matching degree of image T ' and S '；dabBe worth it is smaller, show image T ' and search subgraph (S, a, b)
It is higher with spending；
Step 4: by dabAnd dabCorresponding pixel (a, b) is stored in a two-dimensional array, be denoted as B [r, c] (1≤r≤p, 1
≤c≤p)；
Step 5: repeating step 3 and step 4, until all traversal terminates in neighborhood, find out the smallest similarity measure values of step 4
dmin, by dmihMass center coordinate on image S ' of the corresponding pixel (a, b) as best match position image.
4. method according to claim 1 to 3, which is characterized in that described image, which pre-processes, includes:
Step1: gray processing is carried out to image, converts grayscale image for cromogram；
Step2: gaussian filtering is carried out to grayscale image, the noise spot as caused by environment is removed, obtains Edge preservation and completely compare
Spend high image；
Step3: the image binaryzation that Step2 is obtained separates target with background, increases the specific gravity of useful information, reduces dry
Disturb information specific gravity；
Step4: the image after the binaryzation obtained to Step3 carries out connected domain analysis processing.
5. method according to claim 1 to 4, which is characterized in that the similarity measure values of the similarity measurement method
Calculation formula is formula (1)
Wherein, Hu1 (k) (k=1~7) indicates the 7 invariant moments of template image, and Hu2 (k) (k=1~7) indicates search subgraph
7 invariant moments.
6. a kind of claim 1-5 is any described based on improving the rapid image matching method of HU not bending moment in high-accuracy machinery
Application method in work pieces process detection.
7. application method according to claim 6, which is characterized in that the high-accuracy mechanical workpieces include: semiconductor core
Piece, PCB circuit board, gear rotor, touch display screen, sensor.
8. application method according to claim 7, which is characterized in that the described method includes:
The image of high-accuracy mechanical workpieces is obtained as image to be matched S using industrial camera shooting；
Select that no defect, feature be obvious, workpiece product shooting picture of size qualification is as template from actual production process
Image T；
High-accuracy mechanical workpieces are detected based on the rapid image matching method for improving HU not bending moment using above-mentioned.
9. application method according to claim 8, which is characterized in that the method also includes: before being detected, benefit
High-accuracy mechanical workpieces are positioned and tracked with template image T.
10. application method according to claim 9, which is characterized in that the method also includes: after being detected,
Size detection and/or defect recognition are carried out to high-accuracy mechanical workpieces.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811577536.4A CN109711457A (en) | 2018-12-20 | 2018-12-20 | It is a kind of based on improve the HU not rapid image matching method of bending moment and its application |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811577536.4A CN109711457A (en) | 2018-12-20 | 2018-12-20 | It is a kind of based on improve the HU not rapid image matching method of bending moment and its application |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109711457A true CN109711457A (en) | 2019-05-03 |
Family
ID=66257232
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811577536.4A Pending CN109711457A (en) | 2018-12-20 | 2018-12-20 | It is a kind of based on improve the HU not rapid image matching method of bending moment and its application |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109711457A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110363235A (en) * | 2019-06-29 | 2019-10-22 | 苏州浪潮智能科技有限公司 | A kind of high-definition picture matching process and system |
| CN110426395A (en) * | 2019-07-02 | 2019-11-08 | 广州大学 | A kind of solar energy EL cell silicon chip surface inspecting method and device |
| CN111444964A (en) * | 2020-03-27 | 2020-07-24 | 江南大学 | Multi-target rapid image matching method based on self-adaptive ROI (region of interest) division |
| CN111833239A (en) * | 2020-06-01 | 2020-10-27 | 北京百度网讯科技有限公司 | Image translation method and device, and image translation model training method and device |
| CN114355933A (en) * | 2021-12-31 | 2022-04-15 | 山东新一代信息产业技术研究院有限公司 | Motion control method for robot docking charging pile |
| CN115128598A (en) * | 2022-08-24 | 2022-09-30 | 天津瑞津智能科技有限公司 | Behavior identification method based on fusion of visual perception and radar perception and terminal equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080069453A1 (en) * | 2006-09-05 | 2008-03-20 | Hitachi High-Technologies Corporation | Inspection Apparatus Using Template Matching Method Using Similarity Distribution |
| CN103237156A (en) * | 2013-04-02 | 2013-08-07 | 哈尔滨工业大学 | Modified block matching algorithm applied to electronic image stabilization |
| CN106780571A (en) * | 2016-12-06 | 2017-05-31 | 西安电子科技大学 | A kind of gray level image matching method based on multi-core DSP |
| CN107730510A (en) * | 2017-09-13 | 2018-02-23 | 电子科技大学 | A kind of combination LTCD and Hu not in the COG of bending moment bump dividing method |
-
2018
- 2018-12-20 CN CN201811577536.4A patent/CN109711457A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080069453A1 (en) * | 2006-09-05 | 2008-03-20 | Hitachi High-Technologies Corporation | Inspection Apparatus Using Template Matching Method Using Similarity Distribution |
| CN103237156A (en) * | 2013-04-02 | 2013-08-07 | 哈尔滨工业大学 | Modified block matching algorithm applied to electronic image stabilization |
| CN106780571A (en) * | 2016-12-06 | 2017-05-31 | 西安电子科技大学 | A kind of gray level image matching method based on multi-core DSP |
| CN107730510A (en) * | 2017-09-13 | 2018-02-23 | 电子科技大学 | A kind of combination LTCD and Hu not in the COG of bending moment bump dividing method |
Non-Patent Citations (2)
| Title |
|---|
| 张儒良等: "一种基于Hu不变矩的匹配演化算法", 《西南师范大学学报(自然科学版)》 * |
| 胡敏等: "快速区域质心图像匹配算法", 《电子测量与仪器学报》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110363235A (en) * | 2019-06-29 | 2019-10-22 | 苏州浪潮智能科技有限公司 | A kind of high-definition picture matching process and system |
| CN110363235B (en) * | 2019-06-29 | 2021-08-06 | 苏州浪潮智能科技有限公司 | High-resolution image matching method and system |
| CN110426395A (en) * | 2019-07-02 | 2019-11-08 | 广州大学 | A kind of solar energy EL cell silicon chip surface inspecting method and device |
| CN110426395B (en) * | 2019-07-02 | 2022-02-11 | 广州大学 | Method and device for detecting surface of solar EL battery silicon wafer |
| CN111444964A (en) * | 2020-03-27 | 2020-07-24 | 江南大学 | Multi-target rapid image matching method based on self-adaptive ROI (region of interest) division |
| CN111444964B (en) * | 2020-03-27 | 2023-08-08 | 江南大学 | Multi-target rapid image matching method based on adaptive ROI (region of interest) division |
| CN111833239A (en) * | 2020-06-01 | 2020-10-27 | 北京百度网讯科技有限公司 | Image translation method and device, and image translation model training method and device |
| CN111833239B (en) * | 2020-06-01 | 2023-08-01 | 北京百度网讯科技有限公司 | Image translation method and device and image translation model training method and device |
| CN114355933A (en) * | 2021-12-31 | 2022-04-15 | 山东新一代信息产业技术研究院有限公司 | Motion control method for robot docking charging pile |
| CN115128598A (en) * | 2022-08-24 | 2022-09-30 | 天津瑞津智能科技有限公司 | Behavior identification method based on fusion of visual perception and radar perception and terminal equipment |
| CN115128598B (en) * | 2022-08-24 | 2022-11-15 | 天津瑞津智能科技有限公司 | Behavior identification method based on fusion of visual perception and radar perception and terminal equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109711457A (en) | It is a kind of based on improve the HU not rapid image matching method of bending moment and its application | |
| Zubizarreta et al. | A framework for augmented reality guidance in industry | |
| Wang et al. | Digital image correlation in experimental mechanics and image registration in computer vision: Similarities, differences and complements | |
| CN104981105B (en) | A kind of quickly accurate detection and method for correcting error for obtaining element central and deflection angle | |
| CN103488972B (en) | Fingertip Detection based on depth information | |
| CN108007388A (en) | A kind of turntable angle high precision online measuring method based on machine vision | |
| JP5339065B2 (en) | Object tracking device | |
| CN106936964B (en) | A kind of mobile phone screen angular-point detection method based on Hough transformation template matching | |
| JP6899189B2 (en) | Systems and methods for efficiently scoring probes in images with a vision system | |
| CN107292869B (en) | Image speckle detection method based on anisotropic Gaussian kernel and gradient search | |
| CN109636854A (en) | A kind of augmented reality three-dimensional Tracing Registration method based on LINE-MOD template matching | |
| CN111444964B (en) | Multi-target rapid image matching method based on adaptive ROI (region of interest) division | |
| Yang et al. | Polygon-invariant generalized Hough transform for high-speed vision-based positioning | |
| Wang et al. | Multi-image mosaic with SIFT and vision measurement for microscale structures processed by femtosecond laser | |
| CN107194916A (en) | A kind of vision measurement system of feature based Point matching | |
| Zhang et al. | Automatic crack inspection for concrete bridge bottom surfaces based on machine vision | |
| Long et al. | Cascaded approach to defect location and classification in microelectronic bonded joints: Improved level set and random forest | |
| CN108876819A (en) | A kind of three-dimensional multi-target tracking algorithm based on the more Bernoulli Jacob's filtering of Poisson | |
| Li et al. | Deep learning-based interference fringes detection using convolutional neural network | |
| Dong et al. | Superpixel-based local features for image matching | |
| CN112150541A (en) | Multi-LED wafer positioning algorithm | |
| AUGMENTED | Grayscale image enhancement for enhancing features detection in marker-less augmented reality technology | |
| Sun et al. | Precision work-piece detection and measurement combining top-down and bottom-up saliency | |
| Bui et al. | A texture-based local soft voting method for vanishing point detection from a single road image | |
| CN113689397A (en) | Workpiece circular hole feature detection method and workpiece circular hole feature detection device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190503 |