Abstract
Visual odometry (VO) refers to incremental estimation of the motion state of an agent (e.g., vehicle and robot) by using image information, and is a key component of modern localization and navigation systems. Addressing the monocular VO problem, this paper presents a novel end-to-end network for estimation of camera ego-motion. The network learns the latent subspace of optical flow (OF) and models sequential dynamics so that the motion estimation is constrained by the relations between sequential images. We compute the OF field of consecutive images and extract the latent OF representation in a self-encoding manner. A Recurrent Neural Network is then followed to examine the OF changes, i.e., to conduct sequential learning. The extracted sequential OF subspace is used to compute the regression of the 6-dimensional pose vector. We derive three models with different network structures and different training schemes: LS-CNN-VO, LS-AE-VO, and LS-RCNN-VO. Particularly, we separately train the encoder in an unsupervised manner. By this means, we avoid non-convergence during the training of the whole network and allow more generalized and effective feature representation. Substantial experiments have been conducted on KITTI and Malaga datasets, and the results demonstrate that our LS-RCNN-VO outperforms the existing learning-based VO approaches.
Highlights
Vision-based ego-motion estimation, termed as Visual odometry (VO), is the process of estimating the ego-motion of an agent using the input of a single or multiple cameras attached to it
We evaluated them according to the KITTI VO/simultaneous localization and mapping (SLAM) evaluation metrics defined in [37], i.e., the average translation error (ATE) and the average rotational error (ARE)
Leveraging the power of deep Recurrent-Convolutional Neural Network (CNN), this new paradigm learns a lowerdimensional optical flow (OF) space and models sequential dynamics
Summary
Vision-based ego-motion estimation, termed as VO, is the process of estimating the ego-motion of an agent (e.g., vehicle and robot) using the input of a single or multiple cameras attached to it. Classic geometry-based VO approaches rely on the geometric constraints extracted from imagery for pose estimation They typically consist of a complicated pipeline including camera calibration, feature detection, feature matching (or tracking), outlier rejection (e.g., RANSAC), motion estimation, scale estimation, and local optimization (Bundle Adjustment) [7,8,9]. The APR approaches extract the high-dimensional features from a single image using a base convolutional neural network (CNN) such as VGG or ResNet and regress these features to the absolute camera pose relative to the world coordinate through a fully connected layer. The RPR approaches estimate the pose of a test image relative to one or more training images rather than in absolute scene coordinates They usually stack two consecutive images as input, extract relative geometric features between them, and regress the relative camera pose using a trained CNN.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
