基于视觉目标匹配的场景识别
|
摘要
场景识别是模式识别重要的研究方向,借助计算机在场景中获取模板目标,在视觉定位、智能导航、旅游向导、增强现实等方面应用广泛。本文通过实验对比常用的特征提取方法,分析了各个特征的优缺点,使用Surf的特征点。将场景特征建立KD树,对模板特征点进行k近邻搜索匹配,针对因为图像噪声干扰而造成的错误匹配的问题,提出了基于空间一致性的筛选方法。接着提出了形状模型投票机制,投选出目标在场景图片中的中心点以及目标方框的位置。实验结果表明,本文的场景识别方法在速度和精度上都可以到达实时系统的要求。
Scene recognition is an important research direction of pattern recognition. With the help of computer to obtain template targets in the scene, it is widely used in visual positioning, intelligent navigation, travel guide, and augmented reality. This paper analyzes the advantages and disadvantages of each feature by comparing the commonly used feature extraction methods with experiments, and uses the feature points of Surf. The scene feature is built into the KD tree, and the k-nearest neighbor search is matched for the template feature points. A spatial consistency-based screening method is proposed for the problem of mismatching caused by image noise interference. Then, the shape model voting mechanism is proposed, and the center point of the target in the scene picture and the position of the target box are selected. The experimental results show that the scene recognition method in this paper can reach the requirements of real-time system in both speed and precision.
Scene recognition is an important research direction of pattern recognition. With the help of computer to obtain template targets in the scene, it is widely used in visual positioning, intelligent navigation, travel guide, and augmented reality. This paper analyzes the advantages and disadvantages of each feature by comparing the commonly used feature extraction methods with experiments, and uses the feature points of Surf. The scene feature is built into the KD tree, and the k-nearest neighbor search is matched for the template feature points. A spatial consistency-based screening method is proposed for the problem of mismatching caused by image noise interference. Then, the shape model voting mechanism is proposed, and the center point of the target in the scene picture and the position of the target box are selected. The experimental results show that the scene recognition method in this paper can reach the requirements of real-time system in both speed and precision.
引文
[1] Bagrowski G, Luckner M. Comparison of corner detectors for revolving objects matching task[M]//Artificial Intelligence and Soft Computing. Springer Berlin Heidelberg, 2012.
[2] Alcantarilla P, Nuevo J, Bartoli A. FAST Explicit Diffusion for Accelerated Features in Nonlinear Scale Spaces[C].British Machine Vision Conference. 2013, 1-11.
[3] Althwarin F, Wang C, et al. Improved SIFT-features matching for object recognition[C]//International ConferenceonVisionsofComputerSciense:Bcs Internation Academic Conference. British Computer Society, 2008:178-190.
[4] Zhang X Q, Men T, Liu C, Yang J, Infrared andVisibleImagesRegistrationUsingBEMD andMI[C]//IEEEInternationalConferenceon ComputerScienseandInformationTechnology.Chengdu, 2010:644-647
[5]WeiY,TaoL.EfficientHistogram-Based Sliding Window[C]//IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, San Francisco, 2010:3003-3010
[6] Chen Shu-qiao.A Corner Matching Algorithm based on Harris Operator[C].//2010 2nd International Conference on Information Engineering and Computer Science.v.2.Department of Tourism, Resources and Environment,Zaozhuang University,2010:1256-1257.
[7] Ma C, Wang G, Ban X, et al. SIFT-based matching algorithm and its application in ear recognition[C]. IEEE International Congress on Image and Signal Processing,Biomedical Engineering and Informatics. 2017, 691-695.
[8] Bay H. SURF:Speeded-Up Robust Features[C]//European Conference on Computer Vision, Graz, 2006:404-417
[9] Vourvoulakis, John,Kalomiros, John,Lygouras, John.FPGA accelerator for real-time SIFT matching with RANSAC support[J].Microprocessors and microsystems,2017,49(M ar.):105-116.
[10] Sattler T, Leibe B, Kobbelt L. SCRAMSAC:Improving RANSAC’s efficiency with a spatial consistency filter[C]//IEEE, International Conference on Cpmputer Vision. IEEE,2010:2090-2097.
[2] Alcantarilla P, Nuevo J, Bartoli A. FAST Explicit Diffusion for Accelerated Features in Nonlinear Scale Spaces[C].British Machine Vision Conference. 2013, 1-11.
[3] Althwarin F, Wang C, et al. Improved SIFT-features matching for object recognition[C]//International ConferenceonVisionsofComputerSciense:Bcs Internation Academic Conference. British Computer Society, 2008:178-190.
[4] Zhang X Q, Men T, Liu C, Yang J, Infrared andVisibleImagesRegistrationUsingBEMD andMI[C]//IEEEInternationalConferenceon ComputerScienseandInformationTechnology.Chengdu, 2010:644-647
[5]WeiY,TaoL.EfficientHistogram-Based Sliding Window[C]//IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, San Francisco, 2010:3003-3010
[6] Chen Shu-qiao.A Corner Matching Algorithm based on Harris Operator[C].//2010 2nd International Conference on Information Engineering and Computer Science.v.2.Department of Tourism, Resources and Environment,Zaozhuang University,2010:1256-1257.
[7] Ma C, Wang G, Ban X, et al. SIFT-based matching algorithm and its application in ear recognition[C]. IEEE International Congress on Image and Signal Processing,Biomedical Engineering and Informatics. 2017, 691-695.
[8] Bay H. SURF:Speeded-Up Robust Features[C]//European Conference on Computer Vision, Graz, 2006:404-417
[9] Vourvoulakis, John,Kalomiros, John,Lygouras, John.FPGA accelerator for real-time SIFT matching with RANSAC support[J].Microprocessors and microsystems,2017,49(M ar.):105-116.
[10] Sattler T, Leibe B, Kobbelt L. SCRAMSAC:Improving RANSAC’s efficiency with a spatial consistency filter[C]//IEEE, International Conference on Cpmputer Vision. IEEE,2010:2090-2097.