基于像素级别的Kinect深度测量误差补偿方法  

Pixel-by-pixel depth measurement error compensation method for Kinect

作　　者：牛振岐 刘阔[1] 王月敏[1] 高楠[1] 张宗华[1] 

机构地区：[1]河北工业大学机械工程学院,天津300130

出　　处：《光电子．激光》2016年第11期1169-1175,共7页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金项目(61171048;61311130138);河北省应用基础研究计划重点基础研究项目(15961701D);河北省高等学校高层次人才科学研究项目(GCC2014049);深圳大学省部重点实验室开放基金(GD201401);长江学者和创新团队发展计划(IRT1232)资助项目

摘　　要：本文提出了一种基于像素级别的Kinect2.0深度误差补偿方法。首先,利用Kinect2.0传感器在不同的位置处采集一个标准平面。第二,利用最小二乘法将采集的三维数据进行平面拟合,作为理论平面数据。第三,逐像素点地计算Kinect2.0传感器的深度误差,并建立一个三维查找表。第四,通过建立一个三维查找表,在测量过程中通过插值方法计算每个像素点的深度,实现Kinect2.0测量数据深度方向的误差补偿。实验结果表明,所研究的误差补偿方法在Kinect2.0内部参数未知的情况下,能稳定可靠地将Kinect2.0的平均测量误差减小到0.8mm以内。对于Kinect2.0在较高精度场合的应用,本文算法能够使测量精度达到更高。This paper proposes a novel pixel-by-pixel depth error compensation method for Kinect2.0sensor.Firstly,a standard plane is obtained by Kinect2.0sensor in different positions.Secondly,according to the obtained three -dimensional （3D） data of the standard plan at different distances,the fitting plane which is the theoretical plane data is fitted by th e least squares method.Thirdly, calculate the depth error of Kinect2.0sensor pixel by pixel for establishing a 3D look-up table.Fourthly,through building up a 3D look-up table,depth value of each pixel is calculated by inte rpolation to realize the depth error compensation during measurement.The experimental results show that the proposed method can decrease the mean error of Kinect2.0to 0.8mm with the unknown internal parameters of Kinect 2.0.For the high precision application of Kinect2.0,this paper′s algorithm can achieve higher measure ment accuracy.

关 键 词：Kinect2.0传感器 深度误差补偿 三维查找表 

分 类 号：TH741[机械工程—光学工程] TP391[机械工程—仪器科学与技术]