Abstract：A stereo matching algorithm based on deformable convolution is proposed to perform 3D reconstruction of binocular vision.Firstly, the two-dimensional deformable convolution is used to extract the features of the left and right input images.Secondly, the three-dimensional deformable convolution is used to effectively aggregate the relevant features between the two images in the matching cost volume.Finally, a three-stage cascade residual learning method is used to reduce the parameter calculation amount of the matching cost volume, which can meet the real-time requirements of fast matching.According to the principle of the algorithm, the detection of the disparity depth map is completed, and the three-dimensional object is reconstructed through Open3D.The experimental results show that the parameter amount of the algorithm is 0.5×106, the running time is only 0.02s, the generated disparity map has high precision, and the reconstructed 3D effect is good.
［1］陈辉, 杨剑, 黄晓铭, 等. 基于多视图立体视觉的沙堆三维尺寸测量研究 ［J］. 计量学报, 2019, 40 (3): 403-408.
Chen H, Yang J, Huang X M, et al. Research on 3D Measurement of Sand Pile Based on Multi View Stereo Vision ［J］. Acta Metrologica Sinica, 2019, 40 (3): 403-408.
［2］陈华, 张志娟, 刘刚, 等. 融合色彩分割和加权非参数变换的立体匹配 ［J］. 计量学报, 2017, 38 (4): 406-409.
Chen H, Zhang Z J, Liu G, et al. Stereo Matching Based on the Fusion of Color Segmentation and Weighted Nonparametric Transform ［J］. Acta Metrologica Sinica, 2017, 38 (4): 406-409.
［3］Scharstein D, Szeliski R. A taxonomy and evaluation of dense two-frame stereo correspondence algorithms ［J］. International journal of computer vision, 2002, 47 (1): 7-42.
［4］Hirschmuller H. Accurate and efficient stereo processing by semi-global matching and mutual information［C］//IEEE. 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR′05). 2005, 2: 807-814.
［5］bontar J, LeCun Y. Stereo matching by training a convolutional neural network to compare image patches ［J］. Journal of Machine Learning Research, 2016, 17 (1): 2287-2318.
［6］Mayer N, Ilg E, Hausser P, et al. A large dataset to train convolutional networks for disparity, optical flow, and scene flow estimation［C］//IEEE. Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 4040-4048.
［7］畅雅雯, 赵冬青, 单彦虎. 多特征融合和自适应聚合的立体匹配算法研究 ［J］. 计算机工程与应用, 2021, 57 (23): 219-225.
Chang Y W, Zhao D Q, Shan Y H. Research on Stereo Matching Algorithm Based on Multi-feature Fusion and Adaptive Aggregation ［J］. Computer Engineering and Applications, 2021, 57 (23): 219-225.
［8］Kendall A, Martirosyan H, Dasgupta S, et al. End-to-end learning of geometry and context for deep stereo regression［C］//IEEE. Proceedings of the IEEE international conference on computer vision. 2017: 66-75.
［9］Zhang F, Prisacariu V, Yang R, et al. Ga-net: Guided aggregation net for end-to-end stereo matching［C］//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2019: 185-194.
［10］张浩东, 宋嘉菲, 张广慧. 边缘引导特征融合和代价聚合的立体匹配算法 ［J］. 计算机工程与应用, 2022, 49 (10): 1-8.
Zhang H D, Song J F, Zhang G H. Stereo matching algorithm for edge-guided feature fusion and cost aggregation ［J］. Computer Engineering and Applications, 2022, 49 (10): 1-8.
［11］Saikia T, Marrakchi Y, Zela A, et al. Autodispnet: Improving disparity estimation with automl［C］// IEEE. Proceedings of the IEEE/CVF International Conference on Computer Vision. 2019: 1812-1823.
［12］Hartley R, Zisserman A. Multiple view geometry in computer vision［M］. Cambridge university press, 2003.
［13］Ranganathan A. The levenberg-marquardt algorithm ［J］. Tutoral on LM algorithm, 2004, 11 (1): 101-110.
［14］Liu S, De Mello S, Gu J, et al. Learning affinity via spatial propagation networks ［J］. Advances in Neural Information Processing Systems, 2017, 30.
［15］Zhang Z. A flexible new technique for camera calibration ［J］. IEEE Transactions on pattern analysis and machine intelligence, 2000, 22 (11): 1330-1334.
［16］Menze M, Geiger A. Object scene flow for autonomous vehicles［C］//IEEE. Proceedings of the IEEE conference on computer vision and pattern recognition. 2015: 3061-3070.
［17］Luo W, Schwing A G, Urtasun R. Efficient deep learning for stereo matching［C］// IEEE. Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 5695-5703.
［18］Liang Z, Feng Y, Guo Y, et al. Learning for disparity estimation through feature constancy［C］// IEEE. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018: 2811-2820.
2022, Vol. 43 
