SBAS: Salient Bundle Adjustment for Visual SLAM

Abstract

Recently, the philosophy of visual saliency and attention has started to gain popularity in the robotics community. Therefore, this paper aims to mimic this mechanism in SLAM framework by using saliency prediction model. Comparing with traditional SLAM that treated all feature points as equal important in optimization process, we think that the salient feature points should play more important role in optimization process. Therefore, we proposed a saliency model to predict the saliency map, which can capture both scene semantic and geometric information. Then, we proposed Salient Bundle Adjustment by using the value of saliency map as the weight of the feature points in traditional Bundle Adjustment approach. Exhaustive experiments conducted with the state-of-the-art algorithm in KITTI and EuRoc datasets show that our proposed algorithm outperforms existing algorithms in both indoor and outdoor environments. Finally, we will make our saliency dataset and relevant source code open-source for enabling future research.

A Visual SLAM Approach to Deeply Explore the Spatial Constraints in Indoor Environment Based on the Manhattan Hypothesis

“…Planar Bundle Adjustment (PBA) [15] jointly optimizes the pose of the depth sensor and plane parameters for 3D reconstruction to reduce drift and improve the quality of the map, and such an algorithm using the point-to-plane cost significantly reduces the computational cost. With the philosophy of visual saliency and attention starting to gain popularity in the robotics community, Wang et al [16] proposed a saliency model to predict the saliency map and used the value of the saliency map as the weight of the feature points to improve BA. To prominent robust and rich semantic features, the salient bundle adjustment (SBA) [17] algorithm computes the saliency map by considering both geometric and semantic information and uses the value of the saliency prediction map as the weight of the feature points.…”

“…With the philosophy of visual saliency and attention starting to gain popularity in the robotics community, Wang et al [16] proposed a saliency model to predict the saliency map and used the value of the saliency map as the weight of the feature points to improve BA. To prominent robust and rich semantic features, the salient bundle adjustment (SBA) [17] algorithm computes the saliency map by considering both geometric and semantic information and uses the value of the saliency prediction map as the weight of the feature points. PL-SLAM [18] combined both points and line segments and leveraged both points and line segments at all the instances of the process.…”

ELWNet: An Extremely Lightweight Approach for Real-Time Salient Object Detection

“…object detection [7], semantic segmentation [8], simultaneous localization and mapping [9], RGB-D/T processing [10]- [13] and robot navigation [14], and SOD has been demonstrated to effectively improve the accuracy and robustness of these visual tasks. These methods continuously refresh the detection performance through complex network architectures, advanced feature fusion mechanisms, efficient loss functions and the introduction of edge features, but ignore the huge computational costs and are difficult to deploy on devices with limited computing resources.…”

A Reconfigurable Framework for Vehicle Localization in Urban Areas

“…In the case of vehicle localization, the strategies are usually divided into two main groups: optimization techniques and filtering methods. For the former, commonly applied techniques, such as the Bundle Adjustment [ 25 ], are used because of their consistent and accurate results on various scenarios. On the other hand, filtering methods tend to be based mainly on extended Kalman filters (EKF) and its variants thanks to their convergence and consistency [ 2 ].…”