CN109101985A - It is a kind of based on adaptive neighborhood test image mismatch point to elimination method - Google Patents
It is a kind of based on adaptive neighborhood test image mismatch point to elimination method Download PDFInfo
- Publication number
- CN109101985A CN109101985A CN201810560197.2A CN201810560197A CN109101985A CN 109101985 A CN109101985 A CN 109101985A CN 201810560197 A CN201810560197 A CN 201810560197A CN 109101985 A CN109101985 A CN 109101985A
- Authority
- CN
- China
- Prior art keywords
- point
- neighborhood
- feature
- image
- characteristic point
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/757—Matching configurations of points or features
Abstract
The present invention propose it is a kind of based on adaptive neighborhood test Mismatching point to elimination method, comprising: target image is pre-processed;To treated, image passes through SIFT algorithm extraction characteristic point, and generates Feature Descriptor;The distance between all Feature Descriptors of two images is calculated, characteristic point of the same name is determined by arest neighbors and time neighbour's ratio method, realizes Feature Points Matching;The Mismatching point for being concentrated matching double points by the Mismatching point elimination method tested based on adaptive neighborhood is to rejecting.The present invention capable of only rejecting Mismatching point to higher arest neighbors and time neighbour's rate threshold without will be deleted correct matching double points, is cooperated, can greatly improve matching double points quantity under the premise of same accuracy by carrying out neighborhood test to characteristic point.Experiment shows that the method for the present invention accuracy and number of matches achieve excellent effect.
Description
Technical field
The invention belongs to technical field of image processing, it is related to Feature Points Matching and error hiding occluding technique in three-dimensional reconstruction,
More specifically, be related to it is a kind of based on adaptive neighborhood test image mismatch point to elimination method.
Background technique
Feature Points Matching is the basis much applied in computer vision field, in three-dimensional reconstruction, scene Recognition, map structure
Found a capital and image registration in play an important role.In three-dimensional reconstruction, matching double points are more multiple, and established model is abundanter, matching
Point is more accurate accordingly to more accurate reconstruction model, so Feature Points Matching result directly determines the quality of reconstruction model.
Feature Points Matching process is divided into three steps: characteristic point detection substantially;Construction feature point description;Matching strategy.It is based on
Feature Points Matching, there are many related algorithms in this field, but have certain defect.Such as: the feature that SIFT algorithm extracts
Because of situations such as illumination, noise, Geometrical change when putting and maintain the invariance to rotation, scaling, brightness change, but being matched
It is be easy to cause error hiding, the accuracy of reconstruction model is caused also to decrease, thus before guaranteeing sufficient amount matching double points
It puts, the rejecting of progress error hiding is with regard to meaningful as far as possible.A kind of traditional matching strategy is based on arest neighbors and time neighbour's ratio
The ratio of the matching of value, the i.e. arest neighbors of feature vector and time neighbour are considered same place when being less than threshold value, but due to image line
Manage similar or repeat region presence so that minimum apart from difference between feature vector, institute in this way when rate threshold compared with
There can be more error hiding when big, and when rate threshold is smaller, will appear correct matched accidentally deletion, it is obvious that this method
With quantity, you can't have both at the same time with accuracy.
Summary of the invention
To solve the above problems, the present invention propose a kind of Mismatching point based on adaptive neighborhood test to elimination method,
Reject Mismatching point to while, will not accidentally delete correct matching double points.
In order to achieve the above object, the invention provides the following technical scheme:
It is a kind of based on adaptive neighborhood test image mismatch point to elimination method, comprising the following steps:
Step 1, target image is pre-processed;
Step 2, to treated, image passes through SIFT algorithm extraction characteristic point, and generates Feature Descriptor;
Step 3, the distance between all Feature Descriptors of two images is calculated, is determined by arest neighbors and time neighbour's ratio method
Characteristic point of the same name realizes Feature Points Matching;
Step 4, by the Mismatching point elimination method tested based on adaptive neighborhood by matching double points obtained in step 3
The Mismatching point of concentration is to rejecting.
Further, the step 4 includes following sub-step:
A selects a pair of of characteristic point in the characteristic point of the same name that step 3 obtains, and obtains two characteristic points in two images
Coordinate;
B adaptively delimit neighborhood respectively centered on two characteristic point coordinates, guarantee characteristic point of the same name in two neighborhoods
Number no less than setting number;
C calculates the ratio of the same place number and same place total number that match each other in two neighborhoods, if ratio is less than setting
Threshold value then by this point to rejecting,
D, above-mentioned three step of iteration, until the characteristic point of the same name in step 3 is all processed.
Preferably, the step of adaptively delimiting neighborhood in the step b specifically includes following process:
The feature point number of the same name that the setting initial side length of neighborhood and neighborhood domestic demand include, according to initial edge long value with step 2
Delimit neighborhood centered on middle characteristic point, if the matching characteristic point number for including in neighborhood not enough if by neighborhood side length expansion be twice,
The feature point number of the same name for including in neighborhood is counted again and is judged, until including sufficient amount of characteristic point of the same name in neighborhood.
Preferably, the step 1 specifically includes following process: gray processing processing is carried out to image, then to grayscale image
Carry out Gauss denoising.
Preferably, the step 2 specifically includes following sub-step:
A constructs scale space;
B, scale space extremum extracting are simultaneously accurately positioned extreme point;
C, building 128 dimensional features description.
Preferably, the step 3 specifically includes following sub-step:
A calculates the feature vector distance for being matched all characteristic points in image and matching image, and retain minimum distance with
Secondary minimum distance;
B, carries out nearest neighbor distance and time minimum distance ratio is tested, if being less than threshold value, assert the two of nearest neighbor distance
Characteristic point is same place.
Compared with prior art, the invention has the advantages that and the utility model has the advantages that
The present invention can only reject Mismatching point to without will be deleted correct matching by carrying out neighborhood test to characteristic point
Point pair cooperates higher arest neighbors and time neighbour's rate threshold such as 0.8, can will match under the premise of same accuracy
Point greatly improves quantity.Experiment shows that default threshold 0.6 matches accuracy 95.26% and the method for the present invention accuracy
When 98.41%, the matching double points quantity that the method for the present invention obtains improves 48.5%, reflects the method for the present invention accuracy and matching
Quantity achieves excellent effect.
Detailed description of the invention
Fig. 1 is the Mismatching point provided by the invention based on adaptive neighborhood test to elimination method flow chart.
Specific embodiment
Technical solution provided by the invention is described in detail below with reference to specific embodiment, it should be understood that following specific
Embodiment is only illustrative of the invention and is not intended to limit the scope of the invention.
Mismatching point provided by the invention based on adaptive neighborhood test carries out special elimination method for two images
Sign point matching treatment, process are as shown in Figure 1, comprising the following steps:
Step 1, data prediction is carried out to the two images of acquisition, progress gray processing processing first, then to grayscale image
Carry out Gauss denoising;
Step 2, to treated, two images pass through SIFT algorithm extraction characteristic point respectively, and generate Feature Descriptor,
Specific step is as follows:
A constructs scale space;
B, scale space extremum extracting are simultaneously accurately positioned extreme point;
C, building 128 dimensional features description;
Step 3, Feature Points Matching is carried out, according to Feature Descriptor, to the feature vector v in image Aa, calculate in image B
All feature vectors and vaDistance and find apart from nearest feature vector vbWith secondary nearest feature vector vb', if dist
(va, vb)/dist(va, vb') < T, then remember vaWith vbFor same place, wherein T is threshold value, can be modified as needed;dist(va, vb)
It is characterized vector vaWith vbDistance.All matching same places in two images, i.e. matching double points collection can be obtained by this step.
Step 4, adaptive neighborhood test is carried out to the corresponding dot pair in previous step, rejects Mismatching point pair.
A pair of of same place is selected first and determines its character pair point in the coordinate (x1, y1) and (x2, y2) of two images.
Secondly respectively centered on (x1, y1) and (x2, y2), the neighborhood for adaptively delimiting two centers ensures have in neighborhood
Characteristic point of the same name more than certain amount, i.e., including at least the characteristic point of the same name in the step 3 of N or more in each neighborhood.
Adaptive neighborhood divides specific steps are as follows: the characteristic point of the same name that the setting initial side length of neighborhood and neighborhood domestic demand include
Number delimit neighborhood according to initial edge is long, if the matching characteristic point number for including in neighborhood not enough if the expansion of neighborhood side length is twice,
Feature point number of the same name in neighborhood is counted again and is judged, until including sufficient amount of characteristic point of the same name in neighborhood.
Count in two neighborhoods that the quantity of match point is denoted as m, i.e. feature of the same name in two neighborhoods each other in characteristic point of the same name again
Point is the number of match point each other, calculates the ratio m/ of the feature point number of the same name of match point and feature point number n of the same name each other
N, if ratio is less than the threshold value of setting, this means that this is not right each other to characteristic point most of in feature neighborhood of a point of the same name
It answers, then the namely tested match point of the corresponding dot pair of the centre of neighbourhood at this time must be Mismatching point pair;Then by this to of the same name
Characteristic point is rejected.If more than given threshold then by this point to be added initial matching point to collection.M/n specific value can be adjusted as needed
It is whole, under normal conditions only when matched same place number is much smaller than all feature point numbers of the same name in neighborhood just by matching double points
It rejects.
Remaining corresponding dot pair of iteration is then proceeded to until all characteristic points of the same name that step 3 is found are all processed.
The technical means disclosed in the embodiments of the present invention is not limited only to technological means disclosed in above embodiment, further includes
Technical solution consisting of any combination of the above technical features.It should be pointed out that for those skilled in the art
For, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also considered as
Protection scope of the present invention.
Claims (6)
1. a kind of image mismatch point based on adaptive neighborhood test is to elimination method, which comprises the following steps:
Step 1, target image is pre-processed;
Step 2, to treated, image passes through SIFT algorithm extraction characteristic point, and generates Feature Descriptor;
Step 3, the distance between all Feature Descriptors of two images is calculated, is determined by arest neighbors with secondary neighbour's ratio method of the same name
Characteristic point realizes Feature Points Matching;
Step 4, matching double points obtained in step 3 are concentrated by the Mismatching point elimination method tested based on adaptive neighborhood
Mismatching point to rejecting.
2. the image mismatch point according to claim 1 based on adaptive neighborhood test is to elimination method, feature exists
In the step 4 includes following sub-step:
A selects a pair of of characteristic point in the characteristic point of the same name that step 3 obtains, and obtains coordinate of two characteristic points in two images;
B adaptively delimit neighborhood respectively centered on two characteristic point coordinates, guarantee that feature point number of the same name is not in two neighborhoods
Less than setting number;
C calculates the ratio of the same place number and same place total number that match each other in two neighborhoods, if ratio is less than given threshold
Then by this point to rejecting,
D, above-mentioned three step of iteration, until the characteristic point of the same name in step 3 is all processed.
3. the image mismatch point according to claim 2 based on adaptive neighborhood test is to elimination method, feature exists
In the step of adaptively delimiting neighborhood in the step b specifically includes following process:
The feature point number of the same name that the setting initial side length of neighborhood and neighborhood domestic demand include, first, in accordance with initial edge long value with step 2
Delimit neighborhood centered on middle characteristic point, if the matching characteristic point number for including in neighborhood not enough if by neighborhood side length expansion be twice,
The feature point number of the same name for including in neighborhood is counted again and is judged, until including sufficient amount of characteristic point of the same name in neighborhood.
4. the image mismatch point according to claim 1 based on adaptive neighborhood test is to elimination method, feature exists
In the step 1 specifically includes following process: carrying out gray processing processing to image, then carry out Gauss denoising to grayscale image.
5. the image mismatch point according to claim 1 based on adaptive neighborhood test is to elimination method, feature exists
In the step 2 specifically includes following sub-step:
A constructs scale space;
B, scale space extremum extracting are simultaneously accurately positioned extreme point;
C, building 128 dimensional features description.
6. the image mismatch point according to claim 1 based on adaptive neighborhood test is to elimination method, feature exists
In the step 3 specifically includes following sub-step:
A calculates the feature vector distance for being matched all characteristic points in image and matching image, and retains minimum distance and time most
Closely;
B, carries out nearest neighbor distance and time minimum distance ratio is tested, if being less than threshold value, assert two features of nearest neighbor distance
Point is same place.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810560197.2A CN109101985A (en) | 2018-06-03 | 2018-06-03 | It is a kind of based on adaptive neighborhood test image mismatch point to elimination method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810560197.2A CN109101985A (en) | 2018-06-03 | 2018-06-03 | It is a kind of based on adaptive neighborhood test image mismatch point to elimination method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109101985A true CN109101985A (en) | 2018-12-28 |
Family
ID=64796631
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810560197.2A Pending CN109101985A (en) | 2018-06-03 | 2018-06-03 | It is a kind of based on adaptive neighborhood test image mismatch point to elimination method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109101985A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110070110A (en) * | 2019-03-28 | 2019-07-30 | 陕西理工大学 | A kind of adaptive threshold image matching method |
CN112183596A (en) * | 2020-09-21 | 2021-01-05 | 湖北大学 | Linear segment matching method and system combining local grid constraint and geometric constraint |
CN112381765A (en) * | 2020-10-23 | 2021-02-19 | 杭州翔毅科技有限公司 | Equipment detection method, device, equipment and storage medium based on artificial intelligence |
CN112465025A (en) * | 2020-11-26 | 2021-03-09 | 济南大学 | Image feature matching method and system based on neighborhood inference |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100074528A1 (en) * | 2008-09-23 | 2010-03-25 | Microsoft Corporation | Coherent phrase model for efficient image near-duplicate retrieval |
CN103105851A (en) * | 2012-11-07 | 2013-05-15 | 吉林大学 | Kinesthesis teaching control method based on vision sense for remote control of robot |
CN103177468A (en) * | 2013-03-29 | 2013-06-26 | 渤海大学 | Three-dimensional motion object augmented reality registration method based on no marks |
CN104615642A (en) * | 2014-12-17 | 2015-05-13 | 吉林大学 | Space verification wrong matching detection method based on local neighborhood constrains |
-
2018
- 2018-06-03 CN CN201810560197.2A patent/CN109101985A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100074528A1 (en) * | 2008-09-23 | 2010-03-25 | Microsoft Corporation | Coherent phrase model for efficient image near-duplicate retrieval |
CN103105851A (en) * | 2012-11-07 | 2013-05-15 | 吉林大学 | Kinesthesis teaching control method based on vision sense for remote control of robot |
CN103177468A (en) * | 2013-03-29 | 2013-06-26 | 渤海大学 | Three-dimensional motion object augmented reality registration method based on no marks |
CN104615642A (en) * | 2014-12-17 | 2015-05-13 | 吉林大学 | Space verification wrong matching detection method based on local neighborhood constrains |
Non-Patent Citations (2)
Title |
---|
胡松等: "一种基于图像特征点信息的误匹配点剔除改进算法", 《成都信息工程学院学报》 * |
龚涛: "《摄影测量学》", 30 April 2014, 西南交通大学出版社 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110070110A (en) * | 2019-03-28 | 2019-07-30 | 陕西理工大学 | A kind of adaptive threshold image matching method |
CN110070110B (en) * | 2019-03-28 | 2022-10-21 | 陕西理工大学 | Adaptive threshold image matching method |
CN112183596A (en) * | 2020-09-21 | 2021-01-05 | 湖北大学 | Linear segment matching method and system combining local grid constraint and geometric constraint |
CN112381765A (en) * | 2020-10-23 | 2021-02-19 | 杭州翔毅科技有限公司 | Equipment detection method, device, equipment and storage medium based on artificial intelligence |
CN112465025A (en) * | 2020-11-26 | 2021-03-09 | 济南大学 | Image feature matching method and system based on neighborhood inference |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107680054B (en) | Multi-source image fusion method in haze environment | |
CN102360421B (en) | Face identification method and system based on video streaming | |
CN110909690B (en) | Method for detecting occluded face image based on region generation | |
CN109101985A (en) | It is a kind of based on adaptive neighborhood test image mismatch point to elimination method | |
CN109636784B (en) | Image saliency target detection method based on maximum neighborhood and super-pixel segmentation | |
CN110414439B (en) | Anti-blocking pedestrian tracking method based on multi-peak detection | |
CN107705288A (en) | Hazardous gas spillage infrared video detection method under pseudo- target fast-moving strong interferers | |
CN107066933A (en) | A kind of road sign recognition methods and system | |
CN103440644B (en) | A kind of multi-scale image weak edge detection method based on minimum description length | |
CN105493141B (en) | Unstructured road border detection | |
CN110782477A (en) | Moving target rapid detection method based on sequence image and computer vision system | |
CN109446922B (en) | Real-time robust face detection method | |
CN110490913A (en) | Feature based on angle point and the marshalling of single line section describes operator and carries out image matching method | |
CN116110053B (en) | Container surface information detection method based on image recognition | |
CN110245600B (en) | Unmanned aerial vehicle road detection method for self-adaptive initial quick stroke width | |
CN105809651A (en) | Image saliency detection method based on edge non-similarity comparison | |
CN109978848A (en) | Method based on hard exudate in multiple light courcess color constancy model inspection eye fundus image | |
CN106294705A (en) | A kind of batch remote sensing image preprocess method | |
CN111062331B (en) | Image mosaic detection method and device, electronic equipment and storage medium | |
CN104123554A (en) | SIFT image characteristic extraction method based on MMTD | |
CN105184771A (en) | Adaptive moving target detection system and detection method | |
CN108961301A (en) | It is a kind of based on the unsupervised Chaetoceros image partition method classified pixel-by-pixel | |
CN109447036A (en) | A kind of segmentation of image digitization and recognition methods and system | |
CN113205494B (en) | Infrared small target detection method and system based on adaptive scale image block weighting difference measurement | |
CN104268845A (en) | Self-adaptive double local reinforcement method of extreme-value temperature difference short wave infrared image |
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: 20181228 |