摘要
针对TOF相机获取的原始深度数据存在多种误差,导致成像出现畸变和偏差的问题,建立极限学习机(ELM)空间配准模型,对TOF相机深度数据测量过程中非系统性和系统性误差的叠加导致的深度数据偏移进行统一校正,并与基于BP算法建立的空间配准模型以及基于小孔成像原理校正的结果进行对比。实验结果表明,采用ELM算法所建立的ELM空间配准模型大幅减小了TOF相机测量深度数据的误差。在综合实时性、精确性和泛化能力方面,ELM算法比其他两种方法的数据校正效果更优,能更好地复原真实场景中的深度数据,为TOF相机测量的深度数据的校正提供了一种新的思路和方法。
There are many kinds of errors in the original depth data obtained by the TOF camera,leading to the problems of distortion and deviation in imaging.This paper establishes the extreme learning machine(ELM)spatial registration model to uniformly correct the offset of depth data caused by superposition of non-systematic and systematic errors in the depth data measurement process of the TOF camera.The results were compared with the spatial registration model based on BP algorithm and the calibration results based on the pinhole imaging principle.The experimental results show that the ELM spatial registration model established by the ELM algorithm greatly reduces the error of the depth data measured by the TOF camera.Considering real-time,accuracy and generalization capabilities,ELM algorithm has better data correction effect than the other two methods and can better recover the depth data in the real scenes.It provides a new idea and method for the correction of depth data measured by the TOF camera.
作者
王世程
于微波
杨宏韬
Wang Shicheng;Yu Weibo;Yang Hongtao(College of Electrical and Electronic Engineering,Changchun University of Technology,Changchun 130012,Jilin,China)
出处
《计算机应用与软件》
北大核心
2020年第7期43-48,共6页
Computer Applications and Software
基金
吉林省科技发展计划技术攻关项目(20190303099SF)
吉林省省级产业创新专项资金项目(2019C010)
吉林省科技发展计划重点项目(20180201105GX)。
关键词
TOF深度相机
极限学习机
空间配准模型
误差校正
TOF depth camera
Extreme learning machine
Spatial registration model
Error correction