In practical applications, if there are large numbers of dynamic features with a same movement direction in the image, they will seriously affect the estimation results of the visual odometry. To solve the problem, a stereo visual odometry algorithm is proposed using the location of features to decouple the estimation of camera's rotation and translation. The feature points are divided into "far point" and "near point" by the stereoscopic system. Under the framework of RANSAC (random sample consensus), the "far point" is used to estimate the orientation of the visual system, then the "near point" is used to estimate the translation of the camera under the known orientation condition, which is the core of the rotation-translation decoupling algorithm. Using the algorithm, the impact of the nearby moving objects on the visual odometry is reduced by the rotation constraint. The experiment results show that the proposed rotation-translation decoupling estimation algorithm can more effectively eliminate the dynamic features in the actual road than the traditional algorithms which estimate rotation-translation simultaneously. The proposed algorithm shows its stronger anti-interference against dynamic features, which makes it more robust and accurate.
[1] Maimone M, Cheng Y, Matthies L. Two years of visual odometry on the Mars exploration rovers [J]. Journal of Field Robotics, 2007, 24(3): 169-186. [2] Howard A. Real-time stereo visual odometry for autonomous ground vehicles[C]//IEEE/RSJ International Conference on Intelligent Robots and Systems. Piscataway, USA: IEEE, 2008: 3946-3952.[3] Helmick D, Cheng Y, Clouse D, et al. Path following using visual odometry for a Mars rover in high-slip environments[C]// Proceedings of the IEEE Aerospace Conference, vol.2. Piscataway, USA: IEEE, 2004: 772-789.[4] Nister D. An efficient solution to the five-point relative pose problem[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2004, 26(6): 756-770. [5] Vieville T, Clergue E, Facao P. Computation of ego-motion and structure from visual and inertial sensors using the vertical cue[C]//IEEE International Conference on Computer Vision. Piscataway, USA: IEEE, 1993: 591-598.[6] Kukelova Z, Pajdla T. A minimal solution to the autocalibration of radial distortion[C]//IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, USA: IEEE, 2007: 1-7.[7] Byrod M, Kukelova Z, Josephson K, et al. FAST and robust numerical solutions to minimal problems for cameras with radial distortion[C]//IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, USA: IEEE, 2008: 1-8.[8] Bujnak M, Kukelova Z, Pajdla T. A general solution to the P4P problem for camera with unknown focal length[C]//IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, USA: IEEE, 2008: 9-16.[9] Stewenius H, Engels C, Nister D. An efficient minimal solution for infinitesimal camera motion[C]//IEEE Conference Computer Vision and Pattern Recognition. Piscataway, USA: IEEE, 2007: 1-8.[10] Scaramuzza D, Zurich E, Siegwart R. Appearance-guided monocular omnidirectional visual odometry for outdoor ground vehicles[J]. IEEE Transactions on Robotics, 2008, 24(5): 1015- 1026. [11] Scaramuzza D, Fraundorfer F, Siegwart R. Real-time monocular visual odometry for on-road vehicles with 1-point RANSAC[C]//IEEE International Conference on Robotics and Automation. Piscataway, USA: IEEE, 2009: 4293-4299.[12] Hartley R, Zisserman A. Multiple view geometry in computer vision[M]. Cambridge, UK: Cambridge University Press, 2004.[13] Kitt B, Geiger A, Aegahn H. Visual odometry based on stereo image sequences with RANSAC-based outlier rejection scheme[C]// Proceedings of IEEE Intelligent Vehicles Symposium. Piscataway, USA: IEEE, 2010: 486-492.[14] Atsushi S, Yuya T, Kuroda Y. Visual odometry using feature point and ground plane for urban environments[C]//41st International Symposium on Robotics and 6th German Conference on Robotics. Piscataway, USA: IEEE, 2010: 1-8.[15] Ishii A, Sakai A, Mitsuhashi M, et al. Correcting angle of visual odometry system by fusing monocular and stereo methods in untextured dynamic environment[C]//10th International Symposium on Communications and Information Technologies. Piscataway, USA: IEEE, 2010: 950-955.[16] Kitt B, Moosmann F, Stiller C. Moving on to dynamic environments: Visual odometry using feature classification[C]//IEEE/RSJ International Conference on Intelligent Robots and Systems. Piscataway, USA: IEEE, 2010: 5551- 5556.[17] Guizilini V, Ramos F. Visual, odometry learning for unmanned aerial vehicles[C]//IEEE International Conference on Robotics and Automation. Piscataway, USA: IEEE, 2011: 6213-6220.[18] 阮秋琦.数字图像处理学[M].北京:电子工业出版社, 2007. Ruan Q Q. Digital image processing[M]. Beijing: Publishing House of Electronics Industry, 2007.[19] Lowe, David G. Object recognition from local scale-invariant features[C]//IEEE International Conference on Computer Vision. Piscataway, USA: IEEE, 1999: 1150-1157.[20] Bay H, Ess A, Tuytelaars T, et al. SURF: Speeded up robust features[C]//European Conference on Computer Vision. Piscataway, USA: IEEE, 2006: 404-417.[21] Edward R, Tom D. Fusing points and lines for high performance tracking[C]//IEEE International Conference on Computer Vision. Piscataway, USA: IEEE, 2005: 1508-1515.[22] Kitt B, Geiger A, Lategahn H. Visual odometry based on stereo image sequences with RANSAC-based outlier rejection scheme[C]//IEEE Intelligent Vehicles Symposium. Piscataway, USA: IEEE, 2010: 486-492.[23] Tardif J P, Pavlidis Y, Daniilidis K. Monocular visual odometry in urban environments using an omnidirectional camera[C]//IEEE/RSJ International Conference on Intelligent Robots and Systems. Piscataway, USA: IEEE, 2008: 2531-2538.[24] Sirtkaya S, Seymen B, Alatan A A. Loosely coupled Kalman filtering for fusion of visual odometry and inertial navigation[C]//IEEE International Conference on Information Fusion. Piscataway, USA: IEEE, 2013: 219-226.