CN110163273B - RANSAC algorithm-based image matching method with genetic factors - Google Patents

RANSAC algorithm-based image matching method with genetic factors Download PDF

Info

Publication number
CN110163273B
CN110163273B CN201910398667.4A CN201910398667A CN110163273B CN 110163273 B CN110163273 B CN 110163273B CN 201910398667 A CN201910398667 A CN 201910398667A CN 110163273 B CN110163273 B CN 110163273B
Authority
CN
China
Prior art keywords
image matching
model
matching model
points
sigma
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910398667.4A
Other languages
Chinese (zh)
Other versions
CN110163273A (en
Inventor
汪建国
田恒源
张运良
刘凌
吴神丽
梁小明
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Unversity of Arts and Science
Original Assignee
Xian Unversity of Arts and Science
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Xian Unversity of Arts and Science filed Critical Xian Unversity of Arts and Science
Priority to CN201910398667.4A priority Critical patent/CN110163273B/en
Publication of CN110163273A publication Critical patent/CN110163273A/en
Application granted granted Critical
Publication of CN110163273B publication Critical patent/CN110163273B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Biology (AREA)
  • Biophysics (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physiology (AREA)
  • Genetics & Genomics (AREA)
  • Biomedical Technology (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses an RANSAC algorithm-based image matching method with genetic factors, which comprises the steps of utilizing an SURF algorithm to calculate and find out characteristic points in an image and matching characteristic points in two continuous images as an image matching alternative sample point set; selecting sample points from the sample point set by using a PURSAC algorithm for image matching, obtaining an initial image matching model and calculating a model quality coefficient; judging different processing steps adopted by the current image matching model according to the quality coefficient; and after iteration for several times, if the quality of the image matching model does not meet the condition all the time, selecting a certain proportion of sample points according to the genetic factor lambda to directly perform model matching, and obtaining a final image matching model. The method effectively reduces the iteration times and improves the efficiency and the reliability of model matching. In the image matching process, the problems that RANSAC and other existing algorithms are poor in real-time performance and robustness and too high in algorithm iteration frequency can be solved.

Description

RANSAC algorithm-based image matching method with genetic factors
Technical Field
The invention belongs to the field of machine vision, and particularly relates to a RANSAC algorithm-based model identification method with genetic factors for image identification and feature point matching.
Background
RANSAC (random sample consensus) is a reference algorithm for model fitting in the presence of outliers (or outliers), and is used for solving the problem of matching image feature points, especially in computer vision, such as visual odometry, motion structure and image inpainting. It is based on the assumption that a subset of samples not contaminated by outliers will construct a correct model and is robust to outlier removal and rough model estimation. It follows a framework of assumptions and verifications: a subset of samples of the model fit is randomly drawn from a given dataset for fitting the model hypothesis, and the model fit is evaluated by calculating the distances of all other samples to the model and constructing a subset of interior points. This assumption and verification loop is repeated until the number of iterations is sufficiently large to achieve a predefined success rate, thereby finding a model constructed from interior points. RANSAC is not reliable nor efficient for cases where the requirements for accuracy and real-time are high, especially for cases where the ratio of outliers is high. The method has serious limitations in the aspects of model search iteration termination, model fitting reliability, quality indexes and the like.
To improve RANSAC, many variant algorithms of RANSAC are proposed. The Torr and Zisserman's MLESAC (maximum likelihood estimation sample consensus) uses the same sampling strategy as RANSAC to generate the hypothetical solution, but selects the solution to maximize likelihood, rather than just introduce inliers. Chum and Matas propose locally optimized ransac (lo-ransac) and optimal random ransac to improve the hypothesis generation step, using only the set of interior points of the best model. Nicett proposes a predictive RANSAC for real-time applications, which guarantees that the algorithm terminates within a certain time. Capel developed a statistical test for RANSAC, allowing the scoring process to be terminated early and saving computational costs. Wang and Luo introduced pursa (objective sample consensus) to avoid stochastic model search in RANSAC using geometric analysis of the identification information from the data set and noise model relationships. Uncertainty RANSAC takes the uncertainty of the sample as an outlier to reduce the number of iterations. A deterministic RANSAC method was also introduced to estimate the correct probability of a match.
Although many RANSAC-improving algorithms are proposed, they are still basically on the tracks set by RANSAC and have some common weaknesses. Since the ratio of outliers is unknown, the number of inliers used in RANSAC and its variants is not a true indicator for evaluating the quality of the model fit. They lack a reliable criterion for terminating the iterative model search process. Often their modeling solutions are inaccurate, requiring a large number of unnecessary iterations if an optimal solution is desired. Furthermore, it is assumed that the generated subset selection does not take full advantage of the information generated during the iterative model search. Their iterative model search process only maintains the best model attempt so far, but does not derive clues from all other attempts.
Disclosure of Invention
In order to solve the above-mentioned defects in the prior art, the present invention aims to provide an image matching method with genetic elements based on the RANSAC algorithm. And the method is used for final image matching by fully utilizing effective information in the model of each iteration. The method greatly reduces the iteration times and improves the algorithm efficiency while improving the accuracy. The invention efficiently and accurately solves the problems of matching points of the camera and the calculation of the basic matrix in the stereoscopic vision field by computer vision. By analyzing the quality of the model fit, using the information provided in each model fit attempt, it is then decided that their interior points tend to be included or excluded in the data set when selecting the next subset sample for the model hypothesis.
The invention is realized by the following technical scheme.
An RANSAC algorithm-based image matching method with genetic factors comprises the following steps:
1) calculating and finding out feature points in the images by utilizing an SURF algorithm, and finding out mutually matched feature points in two continuous images as an alternative sample point set for image matching;
2) performing image matching on the selected sample points by using a sample point selection method in a PURSAC algorithm through an iterative least square algorithm to obtain an initial image matching model, wherein the initial matching model is a 3-order square matrix, and calculating a quality coefficient sigma of the initial image matching model;
3) all points in the sample point set are assigned with a genetic factor lambda, and different processing steps adopted by the current image matching model are judged according to the quality coefficient sigma value:
if the quality coefficient sigma value meeting the condition is obtained, a final image matching model is obtained; otherwise, continuing iteration and correcting the genetic factor lambda of the point in the current image matching model;
4) after iteration for several times, if the quality of the image matching model does not meet the condition all the time, selecting a certain proportion of sample points according to the genetic factor lambda to directly carry out image matching model matching, and obtaining a final image matching model.
Further, in the step 2), a sample point selection method in the pursa algorithm is used, that is, a sample point for image matching is selected from the candidate sample point set by using a distance threshold between set sample points.
Further, in the step 2), the process of calculating the quality coefficient σ of the initial image matching model is as follows:
1) setting cycle number N;
2) in the cycle times, the model with the largest number of internal points is always taken as the optimal model;
3) and (3) taking the obtained optimal model as an initial image matching model, and calculating a quality coefficient according to a formula (1).
Further, in the step 2), the quality coefficient σ value satisfying the condition is:
setting the interval of the quality coefficient sigmaHSet to a quality coefficient high threshold, σLSet to a quality coefficient low threshold, σDSetting as a final quality coefficient threshold;
when sigma < sigmaLThe model quality is poor; when σ > σHThe image matching model has better quality; when sigma is at sigmaHAnd σLWhen the area is in the middle, the quality of the image matching model is general; when σ > σDAnd when the image matching model reaches the standard, the image matching model can be used as a final image matching model to be output.
Further, in the step 3), the final matching model is obtained by using a weighted iterative least square method.
Further, in the step 3), the step of correcting the genetic factor λ of the current image matching model inner point is as follows:
a) if σ < σLAll sample points in the image matching model are determined as outer points, and a value for reducing lambda is calculated by using a formula (2);
7 points which are modeled are removed from the alternative sample points, and the remaining alternative points are subjected to initial image matching model modeling by using PURSAC and an iterative least square method again;
b) if σ > σHObtaining an image matching model by using a weighted iterative least square method, and recalculating sigma according to a formula;
c) if σ > σDDetermining a final image matching model, otherwise, providing a lambda value according to a formula, and randomly selecting 7 sample points to perform image matching model modeling;
d) if σ isL＜σ＜σHThe value of λ is adjusted using equation (3).
Further, in the step 4), 60% of sample points are selected according to the genetic factor lambda to directly perform image matching model matching, and the model is optimized by using a weighted iterative least square method to obtain a final image matching model.
Due to the adoption of the technical scheme, the invention has the following beneficial effects:
compared with the conventional RANSAC and derivative algorithm and technology thereof, the model matching efficiency is greatly improved; the correctness of model matching can be accurately judged; and the accuracy and precision of model matching are improved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the principles of the invention:
FIG. 1 is a flowchart of a method for identifying a model with genetic elements based on RANSAC algorithm.
Fig. 2(a) and (b) are sample point sets selected after feature points are extracted and matched respectively;
fig. 3(a) and (b) are displays of the obtained matching model in image matching, respectively.
Detailed Description
The present invention will now be described in detail with reference to the drawings and specific embodiments, wherein the exemplary embodiments and descriptions of the present invention are provided to explain the present invention without limiting the invention thereto.
As shown in fig. 1, the image matching method with genetic elements based on RANSAC algorithm of the present invention specifically includes the following steps:
the method comprises the following steps: calculating and finding out feature points in the images by utilizing an SURF algorithm, and finding out mutually matched feature points in two continuous images as an alternative sample point set for image matching; calculating the characteristic points of the images by using an SURF algorithm, matching the characteristic points in the two images, and recording the information of the characteristic points;
step two: and obtaining an initial matching model of the image by using a PURSAC algorithm, wherein the initial matching model is a 3-order square matrix. Calculating a quality coefficient sigma of the initial matching model for measuring the estimated quality of the current model;
the process of calculating the quality coefficient sigma of the initial image matching model is as follows:
taking the obtained optimal model as an initial image matching model, and calculating a quality coefficient sigma according to a formula (1):
Figure BDA0002059009150000051
wherein: n is a radical ofτAnd N2τThe number of inliers when the inlier threshold is τ and 2 τ, respectively; τ is the point-to-model distance.
Step three: all points in the sample point set are assigned a genetic factor λ to indicate their likelihood of being interior points of the correct image matching model, and different processing steps are taken for the current image matching model according to the quality coefficient σ value:
if the quality coefficient sigma value meeting the condition is obtained, a proper image matching model is found, and then a final image matching model is obtained; otherwise, iteration is continued and the genetic factor lambda of the point in the current image matching model is corrected, so that reference information is provided for the selection of the next model sample point, and the success rate of model selection is improved;
wherein the quality coefficient sigma value satisfying the condition is:
setting the interval of the quality coefficient sigmaHSet to a quality coefficient high threshold, σLSet to a quality coefficient low threshold, σDSetting as a final quality coefficient threshold;
when sigma < sigmaLThe model quality is shown to be poor; when σ > σHThe model quality is better; when at σHAnd σLIn the middle area, the model quality is general; when σ > σDAnd (4) when the model quality reaches the standard, outputting the model as a final model, and recovering the values of all sample genetic factors lambda. The quality coefficient σ is also used to adjust the genetic factor λ value of the inner point, and when the sample point is selected next time, it is determined whether to include or exclude the genetic factor λ value in the sample point set according to the genetic factor λ value, which can effectively reduce the abnormal value ratio in the sample point set. Interior points are points where the model is within a certain threshold. Sample points with higher values of the genetic factor lambda are preferentially selected, while samples with lower values of the genetic factor lambda tend to be excluded.
And the final matching model is a 3-order square matrix and is obtained by using a weighted iterative least square method.
The method comprises the following steps of correcting a genetic factor lambda of a point in a current image matching model:
a)σ＜σLand (3) setting all sample points in the model as outer points, and calculating by using the formula (2):
Figure BDA0002059009150000061
where t is the number of model matches, λt-1After t-1 model matching, the genetic factor value of the sample point;
reducing the value of lambda, removing 7 points utilizing modeling from the alternative sample points, and reusing PURSAC and an iterative least square method for initial image matching model modeling of the remaining alternative points;
b) if σ > σHObtaining an image matching model by using a weighted iterative least square method, and recalculating sigma according to formula (1):
Figure BDA0002059009150000062
c) if σ > σDDetermining a final image matching model, otherwise, providing a lambda value according to a formula 3), and randomly selecting 7 sample points to perform image matching model modeling:
Figure BDA0002059009150000063
d) if σ isL＜σ＜σHThe value of λ is adjusted using equation (3).
Step four: after iteration is carried out for a plurality of times, if the quality of the image matching model does not meet the condition all the time, a certain proportion (such as 60%) of sample points are selected according to the genetic factor lambda to directly carry out image matching model matching, and the obtained model is used for obtaining a final image matching model by using a weighted iterative least square method.
The following is a detailed description of an example of the positioning of the driving route of a vehicle. In this example, the driving path position of the vehicle is determined by matching the image information acquired continuously during the driving of the vehicle.
S1, extracting feature points by using SURF, and finding a sample point set through feature point matching;
s2, establishing an initial matching model by using the PURSAC; the initial matching model is a 3-order square matrix:
calculating the sigma: after obtaining the model, using the formula
Figure BDA0002059009150000071
The value of σ is calculated.
Wherein: n is a radical ofτAnd N2τThe number of inliers when the inlier threshold is τ and 2 τ, respectively; τ is the distance from the point to the model; setting upσLIs 2.8, σHIs 3.5, σDIs 4.5.
S3, taking different steps according to the sigma value, specifically as follows:
a) when sigma < sigmaLThen, all sample points in the model are determined as exterior points, and a formula is utilized:
Figure BDA0002059009150000072
reducing the value of lambda, removing 7 points utilizing modeling from the alternative sample points, and reusing PURSAC and an iterative least square method for initial modeling of the remaining alternative points;
b) when σ > σHThen, a model is obtained by using a weighted iterative least square method and is according to a formula
Figure BDA0002059009150000073
Recalculating sigma;
c) when σ > σDWhen, determining the final model, otherwise, according to the formula
Figure BDA0002059009150000074
Providing a lambda value for modeling again, wherein 7 sample points are randomly selected for modeling in a mode of modeling again;
d) when sigma isL＜σ＜σHWhen, using the formula:
Figure BDA0002059009150000081
and (4) properly adjusting the value of the lambda, reselecting a sample point generation model according to the value of the lambda, and repeating the steps 3-4.
S4, in the case of step c), when the number of iterations reaches the threshold, selecting a sample point with a high λ value for model matching and obtaining an optimal solution specifically includes: when the iteration times reach 30 times, according to the genetic factor lambda value of the sample point, the first 60% of samples are selected as the sample point to directly generate a model, and a final solution is obtained by using a weighted iteration least square method. And if the final result still does not meet the quality standard, the matching is considered to fail.
The feasibility of the algorithm is verified by using the obtained matching model to display in image matching after extracting the feature points from the images (a) and (b) of the images 2 and (3) and (b) and matching the feature points.
Table 1 shows the results of comparing the RANSAC algorithm-based model identification method with genetic elements (genac) with the RANSAC algorithm for image matching. As can be seen from table 1, the image recognition method of GESAC has greatly better performance than the conventional RANSAC.
Figure BDA0002059009150000082
The present invention is not limited to the above-mentioned embodiments, and based on the technical solutions disclosed in the present invention, those skilled in the art can make some substitutions and modifications to some technical features without creative efforts according to the disclosed technical contents, and these substitutions and modifications are all within the protection scope of the present invention.

Claims (6)

1. An RANSAC algorithm-based image matching method with genetic factors is characterized by comprising the following steps:
1) calculating and finding out feature points in the images by utilizing an SURF algorithm, and finding out mutually matched feature points in two continuous images as an alternative sample point set for image matching;
2) performing image matching on the selected sample points by using a sample point selection method in a PURSAC algorithm through an iterative least square algorithm to obtain an initial image matching model, wherein the initial matching model is a 3-order square matrix, and calculating a quality coefficient sigma of the initial image matching model for measuring the estimated quality of the current model;
the process of calculating the quality coefficient sigma of the initial image matching model is as follows:
1) setting cycle number N;
2) in the cycle times, the model with the largest number of internal points is always taken as the optimal model;
3) taking the obtained optimal model as an initial image matching model, and calculating a quality coefficient sigma according to a formula (1):
Figure FDA0002759263910000011
wherein: n is a radical ofτAnd N2τThe number of inliers when the inlier threshold is τ and 2 τ, respectively; τ is the distance from the point to the model;
3) all points in the sample point set are assigned a genetic factor lambda to indicate the possibility that they are interior points of the correct image matching model, and the different processing steps taken by the current image matching model are determined according to the quality coefficient sigma value:
if the quality coefficient sigma value meeting the condition is obtained, a final image matching model is obtained; otherwise, continuing iteration and correcting the genetic factor lambda of the point in the current image matching model;
when the sample point is selected next time, determining whether to include or exclude the sample point from the sample point set according to the lambda value of the genetic factor; preferentially selecting sample points with higher values of the genetic factor lambda, and excluding samples with lower values of the genetic factor lambda;
4) after iteration for several times, if the quality of the image matching model does not meet the condition all the time, selecting a certain proportion of sample points according to the genetic factor lambda to directly carry out image matching model matching, and obtaining a final image matching model.
2. The method according to claim 1, wherein in step 2), a sample point selection method in a PURSAC algorithm is used, that is, a sample point for image matching is selected from a candidate sample point set by using a distance threshold between set sample points.
3. The method according to claim 1, wherein in step 3), the quality coefficient σ value satisfying the condition is:
setting the interval of the quality coefficient sigmaHSet to a quality coefficient high threshold, σLSet to a quality coefficient low threshold, σDSetting as a final quality coefficient threshold;
when sigma < sigmaLThe model quality is poor; when σ > σHThe image matching model has better quality; when sigma is at sigmaHAnd σLWhen the area is in the middle, the quality of the image matching model is general; when σ > σDAnd when the image matching model reaches the standard, the image matching model can be used as a final image matching model to be output.
4. The method according to claim 1, wherein in the step 3), the final matching model is obtained by using a weighted iterative least square method.
5. The method according to claim 1, wherein in the step 3), the step of correcting the genetic factor λ of the point in the current image matching model comprises the following steps:
a) if σ < σLAll sample points in the image matching model are defined as outer points, and the value of reducing lambda is calculated by using formula (2):
Figure FDA0002759263910000021
where t is the number of model matches, λt-1After t-1 model matching, the genetic factor value of the sample point;
7 points which are modeled are removed from the alternative sample points, and the remaining alternative points are subjected to initial image matching model modeling by using PURSAC and an iterative least square method again;
b) if σ > σHObtaining an image matching model by using a weighted iterative least square method, and recalculating sigma according to formula (1):
Figure FDA0002759263910000031
c) if σ > σDAnd determining a final image matching model, otherwise, providing a lambda value according to a formula (3), and randomly selecting 7 sample points to perform image matching model modeling:
Figure FDA0002759263910000032
d) if σ isL＜σ＜σHThe value of λ is adjusted using equation (3).
6. The method according to claim 1, wherein in the step 4), 60% of sample points are selected according to the genetic factor λ and are directly subjected to image matching model matching, and the model is optimized by using a weighted iterative least square method to obtain a final image matching model.
CN201910398667.4A 2019-05-14 2019-05-14 RANSAC algorithm-based image matching method with genetic factors Active CN110163273B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910398667.4A CN110163273B (en) 2019-05-14 2019-05-14 RANSAC algorithm-based image matching method with genetic factors

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910398667.4A CN110163273B (en) 2019-05-14 2019-05-14 RANSAC algorithm-based image matching method with genetic factors

Publications (2)

Publication Number Publication Date
CN110163273A CN110163273A (en) 2019-08-23
CN110163273B true CN110163273B (en) 2021-02-12

Family

ID=67634532

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910398667.4A Active CN110163273B (en) 2019-05-14 2019-05-14 RANSAC algorithm-based image matching method with genetic factors

Country Status (1)

Country Link
CN (1) CN110163273B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113137971A (en) * 2021-03-24 2021-07-20 贵州电网有限责任公司 RANSAC improvement method suitable for instant positioning and map construction

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101739675A (en) * 2009-12-11 2010-06-16 重庆邮电大学 Method and device for registration and synthesis of non-deformed images
CN103955931A (en) * 2014-04-29 2014-07-30 江苏物联网研究发展中心 Image matching method and device
CN107704867A (en) * 2017-08-24 2018-02-16 哈尔滨工业大学 Based on the image characteristic point error hiding elimination method for weighing the factor in a kind of vision positioning
CN109410248A (en) * 2018-10-23 2019-03-01 湖南科技大学 A kind of flotation froth motion feature extracting method based on r-K algorithm
CN109741376A (en) * 2018-11-19 2019-05-10 天津津航技术物理研究所 It is a kind of based on improve RANSAC algorithm in, LONG WAVE INFRARED method for registering images

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9972158B2 (en) * 2016-10-01 2018-05-15 Cantaloupe Systems, Inc. Method and device of automatically determining a planogram in vending
CN107146201A (en) * 2017-05-08 2017-09-08 重庆邮电大学 A kind of image split-joint method based on improvement image co-registration
US10482619B2 (en) * 2017-07-27 2019-11-19 AI Incorporated Method and apparatus for combining data to construct a floor plan
CN109086795A (en) * 2018-06-27 2018-12-25 上海理工大学 A kind of accurate elimination method of image mismatch

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101739675A (en) * 2009-12-11 2010-06-16 重庆邮电大学 Method and device for registration and synthesis of non-deformed images
CN103955931A (en) * 2014-04-29 2014-07-30 江苏物联网研究发展中心 Image matching method and device
CN107704867A (en) * 2017-08-24 2018-02-16 哈尔滨工业大学 Based on the image characteristic point error hiding elimination method for weighing the factor in a kind of vision positioning
CN109410248A (en) * 2018-10-23 2019-03-01 湖南科技大学 A kind of flotation froth motion feature extracting method based on r-K algorithm
CN109741376A (en) * 2018-11-19 2019-05-10 天津津航技术物理研究所 It is a kind of based on improve RANSAC algorithm in, LONG WAVE INFRARED method for registering images

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
Matching with PROSAC-progressive sample consensus;Ondrej Chum et al.;《2005 IEEE COMPUTER SOCIETY CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》;20051231;第1卷;全文 *
SIFT Feature Point Matching Based on Improved RANSAC Algorithm;Guangjun Shi et al.;《2013 Fifth International Conference on Intelligent Human-Machine Systems and Cybernetics》;20131231;第474-477页 *
图像匹配中误匹配点检测技术综述;单小军等;《计算机应用研究》;20150930;第32卷(第9期);第2561-2571页 *

Also Published As

Publication number Publication date
CN110163273A (en) 2019-08-23

Similar Documents

Publication Publication Date Title
CN107886129B (en) Mobile robot map closed-loop detection method based on visual word bag
WO2018108129A1 (en) Method and apparatus for use in identifying object type, and electronic device
US8411966B2 (en) Estimation of image relations from point correspondences between images
CN110120064B (en) Depth-related target tracking algorithm based on mutual reinforcement and multi-attention mechanism learning
CN107633226B (en) Human body motion tracking feature processing method
Dang et al. A Feature Matching Method based on the Convolutional Neural Network.
CN110047095A (en) Tracking, device and terminal device based on target detection
CN108399627B (en) Video inter-frame target motion estimation method and device and implementation device
CN113096169B (en) Non-rigid multimode medical image registration model establishing method and application thereof
EP3012781A1 (en) Method and apparatus for extracting feature correspondences from multiple images
CN106569946B (en) Mobile terminal performance test method and system
CN111009005A (en) Scene classification point cloud rough registration method combining geometric information and photometric information
CN112001401A (en) Training model and training method of example segmentation network, and example segmentation network
WO2023125119A1 (en) Spatio-temporal action detection method and apparatus, electronic device and storage medium
CN110163273B (en) RANSAC algorithm-based image matching method with genetic factors
CN110738695B (en) Image feature point mismatching and removing method based on local transformation model
CN113569758B (en) Time sequence action positioning method, system, equipment and medium based on action triplet guidance
CN109344758B (en) Face recognition method based on improved local binary pattern
Dai et al. Eaainet: An element-wise attention network with global affinity information for accurate indoor visual localization
CN112084855B (en) Outlier elimination method for video stream based on improved RANSAC method
CN112258449A (en) Rapid nodule matching method based on nodule characteristics
CN108256578B (en) Gray level image identification method, device, equipment and readable storage medium
CN115908485A (en) Real-time pose tracking method and system for non-cooperative target in space
CN113554685A (en) Method and device for detecting moving target of remote sensing satellite, electronic equipment and storage medium
CN111882543A (en) Cigarette filter stick counting method based on AA R2Unet and HMM

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