人工智能培训

搜索

深度学习培训:使用事件和帧的异步,光度特征跟踪(Asynchronous, Photometric Feature Tracking using Events

[复制链接]
376156679 发表于 2018-7-26 08:53:30 | 显示全部楼层 |阅读模式
376156679 2018-7-26 08:53:30 59 2 显示全部楼层
深度学习培训:使用事件和帧的异步,光度特征跟踪(Asynchronous, Photometric Feature Tracking using Events and Frames)我们提出了一种方法,利用事件相机和标准相机的互补性来跟踪低延迟的视觉特征。事件相机是输出像素级亮度变化的新型传感器,称为“事件”。它们比标准相机具有显着的优点,即非常高的动态范围,没有运动模糊,以及微秒级的延迟。然而,因为相同的场景模式可以根据运动方向产生不同的事件,所以跨时间建立事件对应是具有挑战性的。相比之下,标准相机提供不依赖于运动方向的强度测量(帧)。我们的方法提取帧中的特征,然后使用事件异步跟踪它们,从而开发两种类型数据中的最佳类型:帧提供不依赖于运动方向的光度表示,并且事件提供低延迟更新。与先前基于启发式的工作相比,这是基于最大似然框架内的生成事件模型直接使用原始强度测量的第一原理方法。因此,我们的方法产生的特征轨迹在各种各样的场景中都比现有技术更精确(亚像素精度)和更长。
We present a method that leverages the complementarity of event cameras andstandard cameras to track visual features with low-latency.Event cameras arenovel sensors that output pixel-level brightness changes, called "events".Theyoffer significant advantages over standard cameras, namely a very high dynamicrange, no motion blur, and a latency in the order of microseconds.However,because the same scene pattern can produce different events depending on themotion direction, establishing event correspondences across time ischallenging.By contrast, standard cameras provide intensity measurements(frames) that do not depend on motion direction.Our method extracts featureson frames and subsequently tracks them asynchronously using events, therebyexploiting the best of both types of data: the frames provide a photometricrepresentation that does not depend on motion direction and the events providelow-latency updates.In contrast to previous works, which are based onheuristics, this is the first principled method that uses raw intensitymeasurements directly, based on a generative event model within amaximum-likelihood framework.As a result, our method produces feature tracksthat are both more accurate (subpixel accuracy) and longer than the state ofthe art, across a wide variety of scenes.深度学习培训:使用事件和帧的异步,光度特征跟踪(Asynchronous, Photometric Feature Tracking using Events and Frames) KJZgAcOKCJ08882r.jpg
URL地址:https://arxiv.org/abs/1807.09713     ----pdf下载地址:https://arxiv.org/pdf/1807.09713    ----深度学习培训:使用事件和帧的异步,光度特征跟踪(Asynchronous, Photometric Feature Tracking using Events and Frames)
回复

使用道具 举报

smlqf3 发表于 2018-7-27 01:59:21 | 显示全部楼层
smlqf3 2018-7-27 01:59:21 显示全部楼层
回复 支持 反对

使用道具 举报

smlqf3 发表于 2018-7-27 07:30:33 | 显示全部楼层
smlqf3 2018-7-27 07:30:33 显示全部楼层
回复 支持 反对

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

376156679当前离线
新手上路

查看:59 | 回复:2

快速回复 返回顶部 返回列表