本篇文章主要介绍iTOF的标定内容,鉴于自身水平所限,如有错误,欢迎批评指正。(欢迎进Q群交流:874653199)
一、简介
从iTOF芯片获取到原始数据后,需要将原始数据进行与标定参数进行一系列运算来获取深度图及点云数据。标定参数是指原始数据至深度图这一计算过程中必不可缺的参数。
iTOF的标定主要可分为以下四部分:
1.相机标定;
2.Global and pixel-dependent offset;
3.Cyclic error;
4.Temperature drifts。
二、iTOF标定
2.1 相机标定
TOF模组由一个RX(接收端)和一个TX(发射端)组成,相机标定针对RX端进行,RX端与普通的RGB模组相似,所以TOF模组需要进行相机标定。相机标定使用 张正友相机标定法进行标定,标定板通常使用棋盘格。iTOF相机标定结果通常如下:
fx,fy表示焦距,cx,cy表示像主点坐标,k1,k2表示径向畸变。由于切向畸变,径向畸变k3影响较小,且TOF精度要求没那么高,所以p1,p2,k3都不考虑。
畸变校正公式如下:
2.2 Global and pixel-dependent offset
1.Calculate the distance at Lens model.
Lens model distance
2.Calculate difference between measured depth maps to Lens model depth maps.
Bad data masked out. (Low signal strength, Noisy data and Edges)
Compare measured depth to known reference
3.Combine 4 orientations into single image and fit polynomial surface to data.
Combine and fit polynomial surface
Gradient error model fit
2.3 Cyclic error
Non-sinusoidal signal: harmonics alias due to under sampling(N=4)
Changes in modulation waveform shape and emitted optical power between exposures
Calculate the difference between the true phase and the measured phase while shifting the laser illumination timing with respect to the reference pulse. The error is modeled by fitting a Fourier series function of predefined order.
:measured phase
:true phase
:cyclic error
使用查表法对测量的相位进行校正,得到理想相位。
cyclic error
2.4 Temperature drifts
Effect of temperature on the illumination driver output signal。
Calibration is currently using a linear temperature offset model. The phase depends linearly on multiple temperature sensors:
It is necessary to measure phase versus temperature on subset of cameras of the one batch in oven before calibration.
The coefficients are calculated from linear regression.
温度漂移