...
首页> 外文期刊>American journal of applied sciences >Pedestrian Detection in RGB-D Data Using Deep Autoencoders | Science Publications
【24h】

Pedestrian Detection in RGB-D Data Using Deep Autoencoders | Science Publications

机译:使用深度自动编码器的RGB-D数据中的行人检测|科学出版物

获取原文
           

摘要

> >Recent popularity of RGB-D sensors mostly comes from the factthat RGB-images and depth maps supplement each other in machine vision tasks,such as object detection and recognition. This article addresses a problem ofRGB and depth data fusion for pedestrian detection. We propose pedestriandetection algorithm that involves fusion of outputs of 2D- and 3D-detectorsbased on deep autoencoders. Outputs are fused with neural network classifiertrained using a dataset which entries are represented by pairs ofreconstruction errors of 2D- and 3D-autoencoders. Experimental results showthat fusing outputs almost totally eliminate false accepts (precision is 99.8%)and brings recall to 93.2% when tested on the combined dataset that includes alot of samples with significantly distorted human silhouette. Though we usewalking pedestrians as objects of interest, there are few pedestrian-specificprocessing blocks in this algorithm, so, in general, it can be applied to anytype of objects.
机译: > >最近RGB-D传感器的普及主要是因为RGB图像和深度图在机器视觉任务(例如物体检测和识别)中相互补充。本文解决了用于行人检测的RGB和深度数据融合的问题。我们提出了行人检测算法,该算法涉及基于深度自动编码器的2D和3D检测器输出的融合。将输出与使用数据集训练的神经网络分类器融合,条目由2D和3D自动编码器的重构误差对表示。实验结果表明,融合输出几乎完全消除了错误接受(精度为99.8%),并且在包含大量人体轮廓明显失真的组合数据集进行测试时,召回率达到93.2%。尽管我们将步行的行人用作感兴趣的对象,但是该算法中几乎没有行人专用的处理块,因此,通常可以将其应用于任何类型的对象。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号