Kinect深度传感器误差模型研究

Research on Error Models for the Kinect Depth Sensor

针对现有Kinect误差模型没有考虑像素点位置对误差的影响,通过对Kinect深度传感器分别在x、y、z方向上的误差研究,考虑像素点对误差的影响并提出新的误差函数模型。首先,通过Kinect和激光雷达获得深度方向(z)的误差;水平(x)和垂直(y)误差通过提出的三维棋盘格测量获得。结果表明,Kinect深度传感器的误差与像素点的坐标和深度值有关,分别为x、y、z方向建立了误差函数模型,同时将所提出的模型与现存的误差模型比较,分析表明所提出的误差模型能够更加精确地表征Kinect传感器的误差分布情况。

In view of the problem that the existing Kinect error models doesn＇t take the pixel location into account,by studying the error distribution on the three dimensions,a new error model is proposed.First,the depth directional error is measured through a laser scan and Kinect.Then,horizontal and vertical errors are measured using a 3D checkerboard designed.Results show that the errors have a great relationship with the pixel location and the depth value.Finally,a new error model is represented in each axis.The resulting models are compared against the existing models,and through the comparisons,the proposed model is shown to be more fit to the realities.