ubuntu18.04 OAK-D kalibr 双目+imu联合标定

编辑：OAK中国
首发：oakchina.cn
喜欢的话，请多多👍⭐️✍

▌前言

Hello，大家好，这里是OAK中国，我是助手君。

将OAK相机接入ROS系统，运行slam定位建图导航等算法时，为了提高定位建图的精度和算法的鲁棒性，除了需要得到双目相机的标定参数，很多算法还需要得到相机imu的标定参数，以及imu+双目的联合标定参数。

目前原厂为我们提供的OAK相机只有双目的标定参数。所以我们采用了第三方标定工具kalibr，在ROS系统中完成双目标定，IMU标定，IMU+双目的联合标定。

这也是之前有不少朋友在群里提到的需求，我们折腾了好几周，终于完成了！（应该值得一个三连吧！）

内容可能会不定期更新，最新教程请查看官网。

▌准备

• 环境干净的ubuntu18.04系统
• OAK-D + USB3.0数据线 (或者是其他支持双目+imu的OAK相机)
• 打印标定板(这里使用的是Aprilgrid)， 下载地址
• 标定板配置文件，根据实际打印大小填写到april_6x6.yml

#example for aprilgrid
target_type: 'aprilgrid' #gridtype
tagCols: 6                  #number of apriltags
tagRows: 6                  #number of apriltags
tagSize: 0.022              #size of apriltag, edge to edge [m]
tagSpacing: 0.3             #ratio of space between tags to tagSize
                            #example: tagSize=2m, spacing=0.5m --> tagSpacing=0.25[-]
#example for checkerboard
#  targetType: 'checkerboard' #gridtype
#  targetCols: 6              #number of internal chessboard corners
#  targetRows: 7              #number of internal chessboard corners
#  rowsMetricSize: 0.06       #size of one chessboard square [m]
#  colsMetricSize: 0.06       #size of one chessboard square [m]
1
2
3
4
5
6
7
8
9
10
11
12
13

▌安装相关程序

ROS以及depthai_ros环境

ROS melodic + depthai_ros 教程链接 注意替换ROS版本号

安装标定工具以及依赖

编译kalibr

sudo apt install libsuitesparse-dev
sudo apt install libv4l-dev
1
2

源码地址 ，克隆源码到src目录下，然后在kalibr_ws工作目录下使用catkin_make命令编译

以下是标定时需要的依赖

sudo apt-get install python-igraph
1

安装ceres

按照 官方教程 安装ceres

编译code_utils

sudo apt install libdw-dev
1

源码地址 ,克隆源码到src目录下

在code_utils下面找到sumpixel_test.cpp，修改#include “backward.hpp”为 #include “code_utils/backward.hpp”，然后在kalibr_ws工作目录下使用catkin_make命令编译

编译imu_utils

源码地址 ，克隆源码到src目录下，然后在kalibr_ws工作目录下使用catkin_make命令编译

▌标定过程

采集数据

启动depthai_examples stereo_inertial_node.launch，这里设置的图像输出帧率是4帧，这样可以让计算量小一点。

cd dai_ws
source devel/setup.bash
roslaunch depthai_examples stereo_inertial_node.launch enableRviz:=false depth_aligned:=false stereo_fps:=4
1
2
3

rostopic list
1

使用到以上图中画框的话题

双目

具体录制方法参考 视频

rosbag record /stereo_inertial_publisher/left/image_rect /stereo_inertial_publisher/right/image_rect -O stereo.bag
1

imu

注意：录制imu数据时相机静止，时间要大于2个小时

rosbag record /stereo_inertial_publisher/imu -O imu.bag
1

双目+imu

具体录制方法参考 视频

rosbag record /stereo_inertial_publisher/imu /stereo_inertial_publisher/left/image_rect /stereo_inertial_publisher/right/image_rect -O stereo_imu.bag
1

标定

双目

注意配置文件路径

cd kalibr_ws
source devel/setup.bash
rosrun kalibr kalibr_calibrate_cameras --bag ../stereo.bag --topics /stereo_inertial_publisher/left/image_rect /stereo_inertial_publisher/right/image_rect --models pinhole-radtan pinhole-radtan --target ../OAK_D/april_6x6.yml
1
2
3

标定成功，但还存在误差

输出cam_chain.yaml

cam0:
cam_overlaps: [1]
camera_model: pinhole
distortion_coeffs: [0.029163490534580016, -0.08332352456606604, -0.001703975799425871, 0.006992997733791862]
distortion_model: radtan
intrinsics: [814.3568321116179, 815.0894236334714, 683.8491150709657, 340.5856601059704]
resolution: [1280, 720]
rostopic: /stereo_inertial_publisher/left/image_rect
cam1:
T_cn_cnm1:
- [0.999984764432588, 0.0029265301147959008, 0.004680419231117285, -0.07535219882179961]
- [-0.002904105777816229, 0.9999843073810051, -0.004790737036519648, 0.00019253365167533625]
- [-0.00469436601929102, 0.004777071614390909, 0.9999775710056993, -0.0005437262712103053]
- [0.0, 0.0, 0.0, 1.0]
cam_overlaps: [0]
camera_model: pinhole
distortion_coeffs: [0.013208623179345822, -0.03975319079350225, -0.000241290621102109, 0.008441954380996652]
distortion_model: radtan
intrinsics: [822.7870689725859, 820.5717606544596, 680.346373177241, 343.830359521292]
resolution: [1280, 720]
rostopic: /stereo_inertial_publisher/right/image_rect
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21

imu

oak_d.launch文件内容，放在imu_utils/launch目录下

    
        
        
        
        
        
    

1
2
3
4
5
6
7
8
9

播放录制的imu数据，200倍速播放

roscore
rosbag play -r 200　imu_utils/imu.bag
1
2

启动oak_d.launch

cd kalibr_ws
source devel/setup.bash
roslaunch imu_utils oak_d.launch
1
2
3

输出imu_param.yaml

%YAML:1.0
---
type: IMU
name: oak-d
Gyr:
unit: " rad/s"
avg-axis:
    gyr_n: 5.1659466110355748e-03
    gyr_w: 2.3398926139856830e-05
x-axis:
    gyr_n: 5.9992047089125330e-03
    gyr_w: 3.1741702735079302e-05
y-axis:
    gyr_n: 4.2744238338741254e-03
    gyr_w: 1.5610981890151188e-05
z-axis:
    gyr_n: 5.2242112903200643e-03
    gyr_w: 2.2844093794339994e-05
Acc:
unit: " m/s^2"
avg-axis:
    acc_n: 2.5112339801819636e-02
    acc_w: 5.4434330052722689e-04
x-axis:
    acc_n: 2.3967193951432327e-02
    acc_w: 7.3928376091566368e-04
y-axis:
    acc_n: 2.2637396692944962e-02
    acc_w: 2.7146510845110392e-04
z-axis:
    acc_n: 2.8732428761081619e-02
    acc_w: 6.2228103221491327e-04
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32

双目+imu

需要三个文件：双目+imu的采集数据、根据生成的imu标定结果填写的imu.yml、双目标定结果

imu.yml

#Accelerometers
accelerometer_noise_density: 2.52e-02   #Noise density (continuous-time)
accelerometer_random_walk:   5.44e-04   #Bias random walk

#Gyroscopes
gyroscope_noise_density:     5.16e-03   #Noise density (continuous-time)
gyroscope_random_walk:       2.34e-05   #Bias random walk

rostopic:                    /stereo_inertial_publisher/imu   #the IMU ROS topic
update_rate:                 200.0      #Hz (for discretization of the values above)
1
2
3
4
5
6
7
8
9
10

rosrun kalibr kalibr_calibrate_imu_camera --target april_6x6.yaml --cam stereo_camchain.yaml --imu imu.yaml --bag stereo_imu_oak_d.bag
1

生成如下结果

camchain-imucam.yaml

cam0:
    T_cam_imu:
    - [0.0012741143007174438, -0.999877856617157, -0.015577178159958124, 0.05611936652731515]
    - [0.9997404060564851, 0.00162798902986383, -0.022725979620935255, 0.00285232424992767]
    - [0.0227485632680666, -0.01554417892321791, 0.9996203686254297, 0.007183919123961078]
    - [0.0, 0.0, 0.0, 1.0]
    cam_overlaps: [1]
    camera_model: pinhole
    distortion_coeffs: [0.042268429058025504, -0.08677899665541859, 0.001180600575988654, 0.0022839809291763198]
    distortion_model: radtan
    intrinsics: [833.53142215564, 833.6574846171179, 664.9153795944901, 354.05210186777356]
    resolution: [1280, 720]
    rostopic: /stereo_inertial_publisher/left/image_rect
    timeshift_cam_imu: 0.0008209265944781894
cam1:
    T_cam_imu:
    - [0.003900423265967401, -0.9999238322965412, -0.011709667106495082, -0.017964937584165064]
    - [0.9996593071739112, 0.0042010652101436285, -0.0257608352286262, 0.002875608017511033]
    - [0.025808066160071114, -0.01160519954583952, 0.9995995513527299, 0.011372828202903269]
    - [0.0, 0.0, 0.0, 1.0]
    T_cn_cnm1:
    - [0.999989071372855, 0.002537659366062306, 0.003926502234769603, -0.07411913670569026]
    - [-0.002525590738806025, 0.9999920811766924, -0.003075544720875524, 0.00018713537154106094]
    - [-0.003934275826358857, 0.0030655943717139626, 0.9999875617250811, 0.004401043432455365]
    - [0.0, 0.0, 0.0, 1.0]
    cam_overlaps: [0]
    camera_model: pinhole
    distortion_coeffs: [0.047927948017714446, -0.08886896066637022, 0.001194906027010198, 0.004430080060758361]
    distortion_model: radtan
    intrinsics: [846.0352688270785, 845.0776741744443, 659.390274553411, 356.3040770945268]
    resolution: [1280, 720]
    rostopic: /stereo_inertial_publisher/right/image_rect
    timeshift_cam_imu: 0.0008178786423846188
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33

▌参考资料

https://docs.oakchina.cn/en/latest/
https://www.oakchina.cn/selection-guide/

---

OAK中国
| OpenCV AI Kit在中国区的官方代理商和技术服务商
| 追踪AI技术和产品新动态

戳「+关注」获取最新资讯↗↗