基于深度神经网络的图像匹配特征点检测方法
The Method of Feature Point Detection for Image Matching Based on Deep Neural Network
摘要: 为实现机器人识别导航中图像的准确匹配，提出一种以新型深度神经网络模型作为检测器进行匹配特征点检测的方法。根据数据集搭建深度神经网络模型实现从图像到特征点概率图的映射；将得到的特征图通过非极大抑制提取关键点的位置；最后根据检测到的关键点进行准确匹配，找到最为匹配的图像。实验表明，通过该方法检测到的关键点匹配率高，且通过匹配特征点可以实现准确的图像校正。相比传统的图像匹配方法，深度神经网络模型作为检测器优势显著。
Abstract: In order to achieve accurate image matching in robot recognition navigation, a new method for detecting matching feature points using a new deep neural network as a detector is proposed. Build a deep neural network model based on the data set to implement the mapping from the image to the feature point probability map. Then extract the position of the key feature points through non-maximum suppression of the obtained feature map. Finally, match the detected key feature points accurately to find the most matching image. Experiments show that the method of key feature points detected by deep neural network can make the matching rate high, and further ex-periments have found that it can also achieve accurate image correction. Compared with the traditional method, the method of deep neural network as a detector has significant advantages.
文章引用：朱苗苗, 张克华. 基于深度神经网络的图像匹配特征点检测方法[J]. 计算机科学与应用, 2020, 10(3): 575-582. https://doi.org/10.12677/CSA.2020.103059

参考文献

[1] Barnea, D.I. and Silverman, H.F. (1972) A Class of Algorithms for Fast Digital Image Registration. IEEE Transactions on Computers, 100, 179-186. [Google Scholar] [CrossRef] 
[2] Lucas, B.D. and Kanade, T. (1981) An Iterative Image Registration Technique with an Application to Stereo Vision.
[3] Harris, C.G. and Stephens, M. (1988) A Combined Corner and Edge Detector. Alvey Vision Conference, New York, 10-5244. [Google Scholar] [CrossRef] 
[4] Lowe, D.G. (2004) Distinctive Image Features from Scale-Invariant Key-points. International Journal of Computer Vision, 60, 91-110. [Google Scholar] [CrossRef] 
[5] Bay, H., Ess, A., Tuytelaars, T., et al. (2008) Speed-ed-up Robust Features (SURF). Computer Vision and Image Understanding, 110, 346-359. [Google Scholar] [CrossRef] 
[6] Ke, Y. and Sukthankar, R. (2004) PCA-SIFT: A More Distinctive Representation for Local Image Descriptors. Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, Washington, 27 June-2 July 2004, 2.
[7] 杨飒, 杨春玲. 基于压缩感知与尺度不变特征变换的图像配准算法[J]. 光学学报, 2014, 34(11): 1110001.
[8] 冯政寿, 王美清. 基于Harris与改进SIFT算法的图像匹配算法[J]. 福州大学学报(自然科学版), 2012, 40(2): 37-41.
[9] Mikolajczyk, K. and Schmid, C. (2001) Indexing Based on Scale Invariant Interest Points. Proceedings Eighth IEEE International Conference on Computer Vision, Vancouver, 7-14 July 2001, 525-531. [Google Scholar] [CrossRef] 
[10] Fischer, P., Dosovitskiy, A. and Brox, T. (2014) Descriptor Matching with Convolutional Neural Networks: A Comparison to Sift. arXiv Preprint arXiv:1405.5769.
[11] Melekhov, I., Kannala, J. and Rahtu, E. (2016) Image Patch Matching Using Convolutional Descriptors with Euclidean Distance. Asian Conference on Computer Vision, Taipei, 20-24 November 2016, 638-653. [Google Scholar] [CrossRef] 
[12] Ezeobiejesi, J. and Bhanu, B. (2018) Patch Based Latent Fingerprint Matching Using Deep Learning. 2018 25th IEEE International Conference on Image Processing (ICIP), Athens, 7-10 October 2018, 2017-2021. [Google Scholar] [CrossRef] 
[13] 李源熠. 基于深度学习的图片匹配算法实现[D]: [硕士学位论文]. 北京: 北京交通大学, 2018.
[14] 万尧. 基于深度学习的图像匹配算法研究[D]: [硕士学位论文]. 南昌: 南昌大学, 2019.
[15] Ronneberger, O., Fischer, P. and Brox, T. (2015) U-Net: Convolutional Networks for Biomedical Image Segmentation. International Conference on Medical Image Computing and Computer-Assisted Intervention, Munich, 5-9 October 2015, 234-241. [Google Scholar] [CrossRef] 
[16] Petersen, P. and Voigtlaender, F. (2018) Optimal Approximation of Piecewise Smooth Functions Using Deep ReLU Neural Networks. Neural Networks, 108, 296-330. [Google Scholar] [CrossRef] [PubMed]
[17] Ioffe, S. and Szegedy, C. (2015) Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift. arXiv Preprint arXiv:1502.03167.