基于density-ORB特征的图像特征点匹配算法
|
摘要
针对ORB(Oriented FAST and Rotated BRIEF)算法中的Steer BRIEF描述子只通过比较两个像素点的灰度信息来决定0/1编码,容易产生特征点误匹配现象,本文提出基于像素密度(pixel density)的ORB特征描述子算法,利用两幅图像中相同区域的某一特征点邻域空间内像素密度的相似性原理,通过比较两个像素点的密度信息来决定0/1编码,计算误匹配率,验证了density-ORB算法在图像模糊、压缩、光照变化、视角变化等条件下的鲁棒性.实验结果表明,该算法减少了特征点的误匹配个数,特征点误匹配率比ORB算法降低了2.80%.
The Steer BRIEF descriptor in the ORB algorithm only determines the 0/1 code by comparing the gray information of two pixels,which is easy to produce the feature point mismatch. This paper proposes an improved ORB featuredescriptor algorithm based on pixel density,using the same region in both images. The principle of similarity of the pixeldensity in the neighborhood of the feature points. The 0/1 encoding is determined by comparing the density information oftwo pixels. Through the calculation of the mismatch rate,the robustness of the density-ORB algorithm under the conditionsof image blur,compression,illumination change and viewing angle change is verified.Experiments show that the algorithm reduces the number of mismatched feature points,and the feature point mismatched rate decreased by 2.80%.
The Steer BRIEF descriptor in the ORB algorithm only determines the 0/1 code by comparing the gray information of two pixels,which is easy to produce the feature point mismatch. This paper proposes an improved ORB featuredescriptor algorithm based on pixel density,using the same region in both images. The principle of similarity of the pixeldensity in the neighborhood of the feature points. The 0/1 encoding is determined by comparing the density information oftwo pixels. Through the calculation of the mismatch rate,the robustness of the density-ORB algorithm under the conditionsof image blur,compression,illumination change and viewing angle change is verified.Experiments show that the algorithm reduces the number of mismatched feature points,and the feature point mismatched rate decreased by 2.80%.
引文
[1] Schmid C,Mohr R. Local grayvalue invariants for imageretrieval[J]. IEEE Transactions on Pattern Analysis andMachine Intelligence,1997,19(5):530-535.
[2] Trzcinski T,Lepetit V. Efficient discriminative projectionsfor compact binary descriptors.[C]//European Conferenceon Computer Vision. Florence,Italy:Springer,2012:228-242.
[3] Lowe D G. Distinctive image features from scale-invariantkeypoints[J]. International Journal of Computer Vision,2004,60(2):91-110.
[4] Bay H, Ess A, Tuytelaars T, et al. Speeded-up robustfeatures[J]. Computer Vision and Image Understanding,2008,110(3):346-359.
[5] Mikolajczyk K,Schmid C. A performance evaluation of localdescriptors.[C]//International Conference on Pattern Reco-gnition. Los Alamitos,USA:IEEE Computer Society,2005:1615-1630.
[6] Tola E, Lepetit V, Fua P. DAISY:An efficient densedescriptor applied to wide baseline stereo[J]. IEEE Tran-sactions on Pattern Analysis and Machine Intelligence,2010,32(5):815-30.
[7] Wang Z,Fan B,Wu F. Local intensity order pattern forfeature description.[C]//IEEE International Conference onComputer Vision. Barcelona,Spain:IEEE Computer Society,2011:603-610.
[8] Fan B,Wu F,Hu Z. Rotationally invariant descriptors usingintensity order pooling[J]. IEEE Transactions on PatternAnalysis and Machine Intelligence,2012,34(10):2031-45.
[9] Ziegler A,Christiansen E,Kriegman D,et al. Locally uniformcomparison image descriptor[C]//International Conferenceon Neural Information Processing Systems. Doha,Qatar:Springer,Berlin,Heidelberg,2012:1-9.
[10] Calonder M,Lepetit V,Strecha C,et al. BRIEF:binaryrobust independent elementary features.[C]//EuropeanConference on Computer Vision. Crete,Greece:Springer,2010:778-792.
[11] Calonder M,Lepetit V,Ozuysal M,et al. BRIEF:compu-ting a local binary descriptorvery fast.[J]. IEEE Transac-tions on Pattern Analysis and MachineIntelligence,2012,34(7):1281-1298.
[12] Leutenegger S, Chli M, Siegwart R Y. BRISK:binaryrobust invariant scalable keypoints.[C]//IEEE Interna-tional Conference on Computer Vision. Barcelona,Spain:IEEE Computer Society,2011:2548-2555.
[13] Ke Y,Sukthankar R. PCA-SIFT:a more distinctive repre-sentation for local image descriptors.[C]//Computer Visionand Pattern Recognition. Washington,DC,USA:IEEE Com-puter Society,2004:506-513.
[14] Rublee E,Rabaud V,Konolige K,et al. ORB:An efficientalternative to SIFT or SURF.[C]//IEEE InternationalConference on Computer Vision. Ontario,Canada:IEEEComputer Society,2012:2564-2571.
[15] Babenko B,Dollar P,Belongie S. Task specific local regionmatching.[C]//IEEE International Conference on ComputerVision. Rio de Janeiro,Brazil:IEEE Computer Society,2007:1-8.
[16] Schmid C,Mohr R,Bauckhage C. Evaluation of interestpoint detectors[J]. International Journal of Computer Vision,2000,37(2):151-172.
[17] Rosten E,Drummond T. Machine learning for high-speedcorner detection.[C]//European Conference on ComputerVision. Graz,Austria:Springer,Berlin,Heidelberg,2006:430-443.
[18] Rosten E,Drummond T. Fusing points and lines for highperformance tracking.[C]//IEEE International Conferenceon Computer Vision. Beijing,China:IEEE Computer Society,2005:1508-1515.
[19] Neubeck A,Gool L V. Efficient non-maximum suppression.[C]//International Conference on Pattern Recognition. HongKong,China:IEEE Computer Society,2006:850-855.
[20]高翔,张涛.视觉SLAM十四讲:从理论到实践[M].北京:电子工业出版社,2017.
[21] Li R,Fang L. Cluster sensing superpixel and grouping.[C]//Computer Vision and Pattern Recognition. Las Vegas,USA:IEEE Computer Society,2016:1350-1358.
[2] Trzcinski T,Lepetit V. Efficient discriminative projectionsfor compact binary descriptors.[C]//European Conferenceon Computer Vision. Florence,Italy:Springer,2012:228-242.
[3] Lowe D G. Distinctive image features from scale-invariantkeypoints[J]. International Journal of Computer Vision,2004,60(2):91-110.
[4] Bay H, Ess A, Tuytelaars T, et al. Speeded-up robustfeatures[J]. Computer Vision and Image Understanding,2008,110(3):346-359.
[5] Mikolajczyk K,Schmid C. A performance evaluation of localdescriptors.[C]//International Conference on Pattern Reco-gnition. Los Alamitos,USA:IEEE Computer Society,2005:1615-1630.
[6] Tola E, Lepetit V, Fua P. DAISY:An efficient densedescriptor applied to wide baseline stereo[J]. IEEE Tran-sactions on Pattern Analysis and Machine Intelligence,2010,32(5):815-30.
[7] Wang Z,Fan B,Wu F. Local intensity order pattern forfeature description.[C]//IEEE International Conference onComputer Vision. Barcelona,Spain:IEEE Computer Society,2011:603-610.
[8] Fan B,Wu F,Hu Z. Rotationally invariant descriptors usingintensity order pooling[J]. IEEE Transactions on PatternAnalysis and Machine Intelligence,2012,34(10):2031-45.
[9] Ziegler A,Christiansen E,Kriegman D,et al. Locally uniformcomparison image descriptor[C]//International Conferenceon Neural Information Processing Systems. Doha,Qatar:Springer,Berlin,Heidelberg,2012:1-9.
[10] Calonder M,Lepetit V,Strecha C,et al. BRIEF:binaryrobust independent elementary features.[C]//EuropeanConference on Computer Vision. Crete,Greece:Springer,2010:778-792.
[11] Calonder M,Lepetit V,Ozuysal M,et al. BRIEF:compu-ting a local binary descriptorvery fast.[J]. IEEE Transac-tions on Pattern Analysis and MachineIntelligence,2012,34(7):1281-1298.
[12] Leutenegger S, Chli M, Siegwart R Y. BRISK:binaryrobust invariant scalable keypoints.[C]//IEEE Interna-tional Conference on Computer Vision. Barcelona,Spain:IEEE Computer Society,2011:2548-2555.
[13] Ke Y,Sukthankar R. PCA-SIFT:a more distinctive repre-sentation for local image descriptors.[C]//Computer Visionand Pattern Recognition. Washington,DC,USA:IEEE Com-puter Society,2004:506-513.
[14] Rublee E,Rabaud V,Konolige K,et al. ORB:An efficientalternative to SIFT or SURF.[C]//IEEE InternationalConference on Computer Vision. Ontario,Canada:IEEEComputer Society,2012:2564-2571.
[15] Babenko B,Dollar P,Belongie S. Task specific local regionmatching.[C]//IEEE International Conference on ComputerVision. Rio de Janeiro,Brazil:IEEE Computer Society,2007:1-8.
[16] Schmid C,Mohr R,Bauckhage C. Evaluation of interestpoint detectors[J]. International Journal of Computer Vision,2000,37(2):151-172.
[17] Rosten E,Drummond T. Machine learning for high-speedcorner detection.[C]//European Conference on ComputerVision. Graz,Austria:Springer,Berlin,Heidelberg,2006:430-443.
[18] Rosten E,Drummond T. Fusing points and lines for highperformance tracking.[C]//IEEE International Conferenceon Computer Vision. Beijing,China:IEEE Computer Society,2005:1508-1515.
[19] Neubeck A,Gool L V. Efficient non-maximum suppression.[C]//International Conference on Pattern Recognition. HongKong,China:IEEE Computer Society,2006:850-855.
[20]高翔,张涛.视觉SLAM十四讲:从理论到实践[M].北京:电子工业出版社,2017.
[21] Li R,Fang L. Cluster sensing superpixel and grouping.[C]//Computer Vision and Pattern Recognition. Las Vegas,USA:IEEE Computer Society,2016:1350-1358.