Abstract
Convolutional Network (ConvNet), with its strong image representation ability, has achieved significant progress in the computer vision and robotic fields. In this paper, we propose a visual localization approach based on place recognition that combines the powerful ConvNet features and localized image sequence matching. The image distance matrix is constructed based on the cosine distance of extracted ConvNet features, and then a sequence search technique is applied on this distance matrix for the final visual recognition. To speed up the computational efficiency, the locality sensitive hashing (LSH) method is applied to achieve real-time performances with minimal accuracy degradation. We present extensive experiments on four real world data sets to evaluate each of the specific challenges in visual recognition. A comprehensive performance comparison of different ConvNet layers (each defining a level of features) considering both appearance and illumination changes is conducted. Compared with the traditional approaches based on hand-crafted features and single image matching, the proposed method shows good performances even in the presence of appearance and illumination changes.
Highlights
Visual-based vehicle localization in changing environments plays an important role in Simultaneous Localization and Mapping (SLAM) as well as the Advanced Driver AssistanceSystems (ADAS) [1]
The F1 scores attained with the conv4 layer of Convolutional Network (ConvNet) for the four different data sets are higher than 0.85, which are significantly better than those of Fast Appearance Based Mapping (FAB-MAP) and Sequence Simultaneous Localisation and Mapping (SeqSLAM); (3) for real-time visual localization, a speed-up method is achieved by approximating the cosine distance between features with a hamming distance over bit vectors obtained by Locality Sensitive Hashing (LSH), by using
For visual localization based on place recognition, the recognition rate at a high precision level is a key indicator in reflecting whether the system is robust enough to determine the position under a changing environment
Summary
Visual-based vehicle localization in changing environments plays an important role in Simultaneous Localization and Mapping (SLAM) as well as the Advanced Driver Assistance. Unlike the other deep-learning-based image recognition tasks using high dimensional features [11,13], we combine the robustness of sequence ConvNet features and the high-dimensional data reducing ability of Locality Sensitive Hashing (LSH) to develop an effective visual localization system for the long-term navigation of autonomous driving. The F1 (the harmonic average of the precision and recall) scores attained with the conv layer of ConvNet for the four different data sets are higher than 0.85, which are significantly better than those of Fast Appearance Based Mapping (FAB-MAP) and Sequence Simultaneous Localisation and Mapping (SeqSLAM); (3) for real-time visual localization, a speed-up method is achieved by approximating the cosine distance between features with a hamming distance over bit vectors obtained by Locality Sensitive Hashing (LSH), by using.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have