...
  • 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Optics Communications: A Journal Devoted to the Rapid Publication of Short Contributions in the Field of Optics and Interaction of Light with Matter >Machine learning enhanced optical distance sensor
【24h】

Machine learning enhanced optical distance sensor

机译：机器学习增强光学距离传感器

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Presented for the first time is a machine learning enhanced optical distance sensor. The distance sensor is based on our previously demonstrated distance measurement technique that uses an Electronically Controlled Variable Focus Lens (ECVFL) with a laser source to illuminate a target plane with a controlled optical beam spot. This spot with varying spot sizes is viewed by an off-axis camera and the spot size data is processed to compute the distance. In particular, proposed and demonstrated in this paper is the use of a regularized polynomial regression based supervised machine learning algorithm to enhance the accuracy of the operational sensor. The algorithm uses the acquired features and corresponding labels that are the actual target distance values to train a machine learning model. The optimized training model is trained over a 1000 mm (or 1 m) experimental target distance range. Using the machine learning algorithm produces a training set and testing set distance measurement errors of <0.8 mm and <2.2 mm, respectively. The test measurement error is at least a factor of 4 improvement over our prior sensor demonstration without the use of machine learning. Applications for the proposed sensor include industrial scenario distance sensing where target material specific training models can be generated to realize low <1% measurement error distance measurements. (C) 2017 Elsevier B.V. All rights reserved.
机译：首次提出的是机器学习增强光学距离传感器。距离传感器基于我们先前说明的距离测量技术，其使用具有激光源的电子控制可变聚焦透镜（ECVFL）来照亮具有受控光束点的目标平面。通过偏离轴相机观察具有不同光斑尺寸的光斑，并处理斑点尺寸数据以计算距离。特别地，本文提出和说明的是使用正则化的多项式回归的受监管机器学习算法来提高操作传感器的准确性。该算法使用所获取的特征和相应的标签，该标签是培训机器学习模型的实际目标距离值。优化的训练模型培训超过1000mm（或1米）的实验目标距离范围。使用机器学习算法可以分别产生训练集和测试设定距离测量误差<0.8 mm和<2.2 mm。在未经使用机器学习的情况下，测试测量误差在我们的先前传感器演示中至少有4倍。所提出的传感器的应用包括工业场景距离感测，其中可以产生目标材料特定的训练模型来实现低<1％的测量误差距离测量。 （c）2017年Elsevier B.V.保留所有权利。

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