大工至善|大学至真分享 http://blog.sciencenet.cn/u/lcj2212916

博文

[转载]【雷达与对抗】【2012.08】基于激光雷达的多目标动态跟踪系统建模

已有 1521 次阅读 2019-10-22 10:43 |系统分类:科研笔记|文章来源:转载


本文为美国宾夕法尼亚州立大学(作者:Mengran Gou)的硕士论文,共77页。

 

由于激光传感器的高精度和成本降低,基于激光的目标和特征检测变得越来越普遍。使用激光测量的最大挑战之一是同时跟踪多个移动物体的运动,以及在移动物体可能被暂时遮挡情况下的追踪;例如,相对于传感器位置,一辆车在另一辆车后面运动。在这种情况下,精确的运动预测至关重要

 

本文建立了适用于激光多目标跟踪系统的运动预测模型。在已有文献中,研究了许多不同的方法。但是,它们中的大多数都只包含一组有限的模型,或者模型过于保守,无法考虑跟踪对象的任何以前信息。本文采用一个在线动态模型来近似物体的运动,该模型的优点包括:a)不需要事先了解物体的运动动力学信息;b)可以通过定期更新来处理运动变化。这些动态模型预测通过与已知模拟数据的比较来评估,然后,利用在城市场景中采集的实测距离数据对该方法进行了测试。结果表明,基于动态运动模型的多跟踪系统能够跟踪不同的动态运动,对运动变化具有鲁棒性,并能经常预测被遮挡物体的状态。

 

Laser-based detection of objects and featuresis becoming increasingly common due to the high accuracy of these sensors andtheir dropping costs. One of the biggest challenges in the use of laser-basedmeasurements is to track the motion of multiple moving objects simultaneouslyand in situations where a moving object might be temporarily occluded; forexample, a vehicle moving behind another vehicle relative to the sensorsposition. In such situations, accurate motion predictions are essential. Thisthesis develops motion-prediction models suitable for laser-based multi-objecttracking systems. In the literature, a number of different approaches have beenstudied. However, most of them cover a limited set of models or the approach’smodel is too conservative to take into account any previous information of theobject. In this thesis, an online dynamic model is applied to approximate themotion of the object. The advantages of this model include: a) it does not needthe prior knowledge of the object’s motion dynamics; b) it can deal with motionchanges by updating periodically. These dynamic model predictions are evaluatedby comparison to known, simulated data. Then, the approach is tested by usingfield-measured range data collected in urban scenes. The results show that thedynamic motion model-based multi-tracking system can track different dynamicmotions and be robust to the motion changes, and can often predict the statesof occluded objects.

 

引言

相关工作

系统设计

理论方法

实验结果

结论


更多精彩文章请关注公众号:qrcode_for_gh_60b944f6c215_258.jpg



https://blog.sciencenet.cn/blog-69686-1202986.html

上一篇:[转载]【计算机科学】【2016】基于视觉注意力的动作识别和视频描述
下一篇:[转载]【计算机科学】【2012】基于神经网络的统计语言模型
收藏 IP: 112.31.16.*| 热度|

0

该博文允许注册用户评论 请点击登录 评论 (0 个评论)

数据加载中...
扫一扫,分享此博文

Archiver|手机版|科学网 ( 京ICP备07017567号-12 )

GMT+8, 2024-3-28 18:17

Powered by ScienceNet.cn

Copyright © 2007- 中国科学报社

返回顶部