基于SURF和FREAK的移动终端动态背景运动目标检测
|
摘要
针对使用移动终端检测运动目标时出现的背景偏移,实时性不足等问题,本文提出一种基于Speeded-Up Robust Features(SURF)和Fast Retina Keypoint(FREAK)算法的动态背景补偿方法。首先利用SURF算法检测特征点,接着利用FREAK算法对特征点进行描述,然后对特征点进行汉明距离匹配,最后使用随机抽样一致算法(Random Sample Consensus,RANSAC)剔除误匹配点。设计基于移动终端的背景补偿实验,结果表明,在旋转角度,光照条件和尺寸不同的情况下,该算法都表现出良好的匹配效果以及实时性。
A dynamic background compensation method based on speeded-up robust features(SURF) and fast retina keypoint(FREAK) algorithm is proposed in this paper to solve the problem of background deviation and lack of real time ability when detecting moving target with mobile terminal.Considering the redundancy degree of SURF algorithm in feature description phase and the deficiency of FREAK algorithm in feature detection.the two algorithms are combined to complement each other to get the best results.Firstly,the SURF algorithm is used to detect feature points and the FREAK algorithm is used to describe the feature points.Then hamming distance matching is performed on feature points.and the last random sample consensus(RANSAC) algorithm is used to eliminate mismatched points.The background compensation experiment based on mobile terminal is designed.The results show that under different rotation angles,lighting conditions and sizes,compared with the SURF algorithm,the speed of this algorithm is improved by about four times,and the matching accuracy is basically 95%,which is much higher than that of FREAK algorithm,showing good matching effect and real-time performance.
A dynamic background compensation method based on speeded-up robust features(SURF) and fast retina keypoint(FREAK) algorithm is proposed in this paper to solve the problem of background deviation and lack of real time ability when detecting moving target with mobile terminal.Considering the redundancy degree of SURF algorithm in feature description phase and the deficiency of FREAK algorithm in feature detection.the two algorithms are combined to complement each other to get the best results.Firstly,the SURF algorithm is used to detect feature points and the FREAK algorithm is used to describe the feature points.Then hamming distance matching is performed on feature points.and the last random sample consensus(RANSAC) algorithm is used to eliminate mismatched points.The background compensation experiment based on mobile terminal is designed.The results show that under different rotation angles,lighting conditions and sizes,compared with the SURF algorithm,the speed of this algorithm is improved by about four times,and the matching accuracy is basically 95%,which is much higher than that of FREAK algorithm,showing good matching effect and real-time performance.
引文
[1] Lowe D G.Distinctive image features from scale-invariant keypoints[J].International journal of computer vision,2004,60(2):91-110.
[2] Bay H,Tuytelaars T,Van Gool L.Surf:speeded up robust features[M].Computer Vision-ECCV,2006,404-417.
[3] Jain S,Kumar B L S,Shettigar R.Comparative study on SIFT and SURF face feature descriptors[C].International Conference on Inventive Communication and Computational Technologies,2017,200-205.
[4] Thompson F,Jeyakumar M K.Vector based classification of dermoscopic images using SURF[J].International Journal of Applied Engineering Research,2017,12(8):1758-1764.
[5] Schiffer ü,Sari Z,Müller P.Ragweed detection based on SURF features[J].Tehnicki Vjesnik,2017,24(5):1519-1524.
[6] Cheng X,Hao Q,Xie M.A comprehensive motion estimation technique for the improvement of EIS methods based on the SURF algorithm and Kalman filter[J].Sensors,2016,16(4):486.
[7] Oliveira S A F,Neto A R R,Bezerra F N.A novel genetic algorithms and SURF-based approach for image retargeting[J].Expert Systems with Applications,2016,44(C):332-343.
[8] Kim Y,Jung H.Reconfigurable hardware architecture for faster descriptor extraction in SURF[J].Electronics Letters,2018,54(4):210-212.
[9] Ameen M M,Eleyan A.Score fusion of SIFT & SURF descriptors for face recognition using wavelet transforms[J].International Journal of Image Graphics & Signal Processing,2017,9(10):22-28.
[10] Ortiz R.FREAK:Fast retina keypoint[C]. IEEE Conference on Computer Vision and Pattern Recognition.IEEE Computer Society,2012,510-517.
[11] Karami E,Prasad S,Shehata M.Image matching using SIFT,SURF,BRIEF and ORB:performance comparison for distorted images[J].2017.
[12] Bartolini I,Patella M.Windsurf:the best way to SURF: (and SIFT/BRISK/ORB/FREAK,too)[J].Multimedia Systems,2017(3).
[13] Kashif M,Deserno T M,Haak D,et al.Feature description with SIFT,SURF,BRIEF,BRISK,or FREAK? A general question answered for bone age assessment[J].Computers in Biology & Medicine,2016,68(C):67-75.
[14] Gomez C H,Medathati K,Kornprobst P,et al.Improving FREAK descriptor for image classification[C].International Conference on Computer Vision Systems.Springer-Verlag New York,Inc.,2015,14-23.
[15] Prathap K S V,Jilani S A K,Reddy P R.A real-time image mosaicing using FAST detector and FREAK descriptor[C].International Conference on Wireless Communications,Signal Processing and Networking.IEEE,2018,2413-2418.
[2] Bay H,Tuytelaars T,Van Gool L.Surf:speeded up robust features[M].Computer Vision-ECCV,2006,404-417.
[3] Jain S,Kumar B L S,Shettigar R.Comparative study on SIFT and SURF face feature descriptors[C].International Conference on Inventive Communication and Computational Technologies,2017,200-205.
[4] Thompson F,Jeyakumar M K.Vector based classification of dermoscopic images using SURF[J].International Journal of Applied Engineering Research,2017,12(8):1758-1764.
[5] Schiffer ü,Sari Z,Müller P.Ragweed detection based on SURF features[J].Tehnicki Vjesnik,2017,24(5):1519-1524.
[6] Cheng X,Hao Q,Xie M.A comprehensive motion estimation technique for the improvement of EIS methods based on the SURF algorithm and Kalman filter[J].Sensors,2016,16(4):486.
[7] Oliveira S A F,Neto A R R,Bezerra F N.A novel genetic algorithms and SURF-based approach for image retargeting[J].Expert Systems with Applications,2016,44(C):332-343.
[8] Kim Y,Jung H.Reconfigurable hardware architecture for faster descriptor extraction in SURF[J].Electronics Letters,2018,54(4):210-212.
[9] Ameen M M,Eleyan A.Score fusion of SIFT & SURF descriptors for face recognition using wavelet transforms[J].International Journal of Image Graphics & Signal Processing,2017,9(10):22-28.
[10] Ortiz R.FREAK:Fast retina keypoint[C]. IEEE Conference on Computer Vision and Pattern Recognition.IEEE Computer Society,2012,510-517.
[11] Karami E,Prasad S,Shehata M.Image matching using SIFT,SURF,BRIEF and ORB:performance comparison for distorted images[J].2017.
[12] Bartolini I,Patella M.Windsurf:the best way to SURF: (and SIFT/BRISK/ORB/FREAK,too)[J].Multimedia Systems,2017(3).
[13] Kashif M,Deserno T M,Haak D,et al.Feature description with SIFT,SURF,BRIEF,BRISK,or FREAK? A general question answered for bone age assessment[J].Computers in Biology & Medicine,2016,68(C):67-75.
[14] Gomez C H,Medathati K,Kornprobst P,et al.Improving FREAK descriptor for image classification[C].International Conference on Computer Vision Systems.Springer-Verlag New York,Inc.,2015,14-23.
[15] Prathap K S V,Jilani S A K,Reddy P R.A real-time image mosaicing using FAST detector and FREAK descriptor[C].International Conference on Wireless Communications,Signal Processing and Networking.IEEE,2018,2413-2418.