人工智能培训

搜索

人工智能论文:隐私保护基于图像的本地化(Privacy Preserving Image-Based Localization)

[复制链接]
deshuo 发表于 2019-3-15 13:21:45 | 显示全部楼层 |阅读模式
deshuo 2019-3-15 13:21:45 130 0 显示全部楼层
人工智能论文:隐私保护基于图像的本地化(Privacy Preserving Image-Based Localization)基于图像的定位是许多增强/混合现实(AR / MR)和自主机器人系统的核心组件。当前的定位系统依赖于场景的3D点云的持久存储以实现相机姿态估计,但是这样的数据揭示了潜在敏感的场景信息。这会带来严重的隐私风险,特别是对于许多应用程序而言,3Dmapping是用户可能不完全了解的后台进程。我们提出以下问题:如何避免披露有关捕获的3D场景的机密信息,并允许可靠的相机姿态估计?本文提出了我们称之为隐私基于图像的本地化的第一种解决方案。我们的方法的关键思想是将地图表示从3D点云提升到3D线云。这种新颖的表现方式模糊了基础场景几何,同时提供了足够的几何约束,以实现稳健和准确的6-DOF摄像机估计。对几个数据集和局部化方案进行了广泛的实验,强调了我们提出的方法的高度实用性。
Image-based localization is a core component of many augmented/mixed reality(AR/MR) and autonomous robotic systems.Current localization systems rely onthe persistent storage of 3D point clouds of the scene to enable camera poseestimation, but such data reveals potentially sensitive scene information.Thisgives rise to significant privacy risks, especially as for many applications 3Dmapping is a background process that the user might not be fully aware of.Wepose the following question: How can we avoid disclosing confidentialinformation about the captured 3D scene, and yet allow reliable camera poseestimation?This paper proposes the first solution to what we call privacypreserving image-based localization.The key idea of our approach is to liftthe map representation from a 3D point cloud to a 3D line cloud.This novelrepresentation obfuscates the underlying scene geometry while providingsufficient geometric constraints to enable robust and accurate 6-DOF camerapose estimation.Extensive experiments on several datasets and localizationscenarios underline the high practical relevance of our proposed approach.人工智能论文:隐私保护基于图像的本地化(Privacy Preserving Image-Based Localization) yZ5s5tzf7fE3tF25.jpg
URL地址:https://arxiv.org/abs/1903.05572     ----pdf下载地址:https://arxiv.org/pdf/1903.05572    ----人工智能论文:隐私保护基于图像的本地化(Privacy Preserving Image-Based Localization)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

deshuo当前离线
新手上路

查看:130 | 回复:0

快速回复 返回顶部 返回列表