Person Property Estimation Based on 2D LiDAR Data Using Deep Neural Network
Person Property Estimation Based on 2D LiDAR Data Using Deep Neural Network
复制标题
使用深度神经网络基于 2D LiDAR 数据的人物属性估计
DOI：
10.1007/978-3-030-84522-3_62
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Kobayashi Yoshinori
中科院分区：
文献类型：
--
作者：
Hasan Mahmudul;Goto Riku;Hanawa Junichi;Fukuda Hisato;Kuno Yoshinori;Kobayashi Yoshinori
基于视频的估计在人员识别和跟踪中发挥着非常重要的作用。新技术的出现和计算能力的增强使系统日益强大和准确。随着时间的推移，这些应用中会使用不同的 RGB 和密集相机。基于视频的分析具有攻击性，并且个人身份会被泄露。作为视觉捕捉的替代方案，现在激光雷达以极高的准确性展示了其实力。除了隐私问题之外，关键的自然环境也可以通过激光雷达传感来解决。一些易受影响的场景，如环境中的大雾和烟雾，是通过典型的视觉估计向下执行的。在这项研究中，我们找到了一种基于激光雷达数据来估计一个人的财产的方法，即身高和年龄等。我们将不同的 2D LiDAR 放置在脚踝处并捕捉人的动作。这些距离数据被处理为运动历史图像。我们使用深度神经架构来估计人的属性并取得了显着的准确性。这种基于 2D LiDAR 的估计可以成为关键推理和环境的新途径。此外，计算成本和准确性对传统方法影响很大。
Video-based estimation plays a very significant role in person identification and tracking. The emergence of new technology and increased computational capabilities make the system robust and accurate day by day. Different RGB and dense cameras are used in these applications over time. Video-based analyses are offensive, and individual identity is leaked. As an alternative to visual capturing, now LiDAR shows its credentials with utmost accuracy. Besides privacy issues but critical natural circumstances also can be addressed with LiDAR sensing. Some susceptible scenarios like heavy fog and smoke in the environment are downward performed with typical visual estimation. In this study, we figured out a way of estimating a person's property, i.e., height and age, etc., based on LiDAR data. We placed different 2D LiDARs in ankle levels and captured persons' movements. These distance data are being processed as motion history images. We used deep neural architecture for estimating the properties of a person and achieved significant accuracies. This 2D LiDAR-based estimation can be a new pathway for critical reasoning and circumstances. Furthermore, computational cost and accuracies are very influential over traditional approaches.