Robust and Efficient Relative Pose With a Multi-Camera System for Autonomous Driving in Highly Dynamic Environments

Abstract

This paper studies the relative pose problem for autonomous vehicle driving in highly dynamic and possibly cluttered environments. This is a challenging scenario due to the existence of multiple, large, and independently moving objects in the environment, which often leads to excessive portion of outliers and results in erroneous motion estimation. Existing algorithms cannot cope with such situations well. This paper proposes a new algorithm for relative pose using a multi-camera system with multiple non-overlapping individual cameras. The method works robustly even when the numbers of outliers are overwhelming. By exploiting specific prior knowledge of driving scene we have developed an efficient 4-point algorithm for multi-camera relative pose, which admits analytic solutions by solving a polynomial root-finding equation, and runs extremely fast (at about 0.5µs per root). When the solver is used in combination with RANSAC, we are able to quickly prune unpromising hypotheses, significantly improve the chance of finding inliers. Experiments on synthetic data have validated the performance of the proposed algorithm. Tests on real data further confirm the method's practical relevance.

PCAN: 3D Attention Map Learning Using Contextual Information for Point Cloud Based Retrieval

“…The task of visual localization is a core problem in computer vision, which can be widely applied in augmented reality [24], robot navigation [22,6,39,38], and autonomous driving [7,18]. Visual localization is always defined as follows: Given an image from a scene, the purpose is to predict the location where the photo is taken.…”

Efficient Ego-Motion Estimation for Multi-Camera Systems With Decoupled Rotation and Translation

“…Thus, this makes the ego-motion estimation process simpler and faster. This ideal has been applied in the monocular camera system [13], [17]- [19] and the multi-camera system [20]- [22].…”

Robust Vehicle and Surrounding Environment Dynamic Analysis for Assistive Driving Using Visual-Inertial Measurements

“…Over the past decades, the dramatic improvements on mobile robotics and computer vision techniques greatly enable the popularity of assistive driving, which could liberate the drivers from steering the vehicles in a long term [1]- [3]. The robust and accurate traffic scene analysis is a prerequisite for reliable assistive driving [4], [5]. This is especially important for assistive driving in complex traffic conditions, e.g., during rush hours in downtown areas in extreme weather conditions, so that the autonomous system could yield proper steering commands [6]- [8].…”

“…Both the ego motion and surrounding vehicle motions are estimated theoretically in a unified framework. However, the monocular vision system fails to robustly estimate the surrounding vehicle dynamics due to illumination variance, feature matching failures, dynamic motion over-abundance [4], and image blurrings [13]. For example, during the on-road vehicle motion estimation, if the number of dynamic inliers is more than the number of static inliers, such as the case shown in Fig.…”