关于Realsense D435i运行VINS系列
前言
在SLAM中,主要是以激光SLAM和视觉SLAM为主,激光雷达直接可以获取三维点云坐标信息,所以激光SLAM会比视觉SLAM稳定许多,但是由于激光雷达挺贵的,而相机成本低廉许多,所以视觉SLAM在工业应用中起到很多作用,视觉SLAM的性能也逐渐发展起来,有时精度甚至可以和激光SLAM可媲美。而在视觉SLAM中相机的FOV有限,不像三位机械旋转式激光雷达那样有水平范围为360度,所以当相机旋转比较大或运动较快或曝光度突然变化的时候(总结来说就是场景变化过大而导致特征匹配不上)很容易跟丢而IMU作为廉价的传感器测量线加速度和角速度(当然九轴IMU可以直接测量姿态角),噪声较大,短时间内可以提供比较准确的测量,而相机可以长时间运行,噪声较小,两种传感器原理上基本是一种互补的关系,且两种传感器的成本都可以做到比较低廉且重量很轻,不像激光雷达那么高昂且较重。所以现在的无人机定位方案基本首选视觉惯性里程计方案(Visual Inertial Odometry,VIO),而自动驾驶L4级别中定位方案基本采用激光惯性里程计(Lidar Inertial Odometry,LIO)。
在SLAM中,VIO具有完整理论紧耦合优化的理论方案为港科大的VINS,也是为数不多具有开创性将VO与IMU的优点结合的工作,精华同样在于VO与IMU紧耦合工作上。所以市面上很多商业方案如VR、科研无人机都是采用VINS方案,包括浙大FAST实验室的高飞博士进行运动规划全是采用VINS提供无人机位置姿态信息(师出同门)。在2017年开源VINS-Mono,支持单目相机,可以在线估计camera与IMU的外參,同样可以在iphone手机上运行(他们提供了VINS-Mobile)。后一年开源VINS-Fusion,支持双目相机+IMU。在2021年开源GVINS,将GNSS、VIO都融合到一个系统,即便在比较极端情况下,GVINS稳定性与精度也都由于前两者。
关于VINS-Mono的代码注释与详解:https://blog.csdn.net/qq_41839222/article/details/85793998
(该博文讲解非常详细),当然深蓝学院也有高翔和贺一家关于从零开始手写VIO的课程。
现在用同一个Realsense D435i分别设备演示一下运行VINS-Mono和VINS-Fusion,其中D435i包含有一个RGB成像模块、两个红外成像模块、一个深度成像模块和一个IMU。
D435i运行VINS-Mono
该程序需要在Ubuntu ROS下运行,参见ROS安装,原作者使用的系统是Ubuntu16.04,而我这里的电脑是Ubuntu20.04,有部分的包可能不一样。安装流程如下:
-
安装ROS相关包
sudo apt-get install ros-noetic-cv-bridge ros-noetic-tf ros-noetic-message-filters ros-noetic-image-transport
-
安装ceres,ceres安装教程,其中现在版本是最新的2.1,VINS-Mono貌似不支持这个版本,换成1.14
-
克隆编译
mkdir -p VINS_ws/src cd VINS_ws/src git clone https://github.com/HKUST-Aerial-Robotics/VINS-Mono.git catkin build source devel/setup.bash
其中源码测试都是基于OpenCV3完成的,但是Ubuntu20.04 ROS都是OpenCV4,所以需要改成对应的函数变量名:
原函数名(OpenCV3) 新函数名(OpenCV4) CV_GRAY2RGB
cv::COLOR_GRAY2RGB
CV_BGR2GRAY
cv::COLOR_BGR2GRAY
CV_LOAD_IMAGE_GRAYSCALE
cv::IMREAD_GRAYSCALE
CV_AA
cv::LINE_AA
CV_CALIB_CB_ADAPTIVE_THRESH
cv::CALIB_CB_ADAPTIVE_THRESH
CV_CALIB_CB_NORMALIZE_IMAGE
cv::CALIB_CB_NORMALIZE_IMAGE
CV_CALIB_CB_FILTER_QUADS
cv::CALIB_CB_FILTER_QUADS
CV_CALIB_CB_FAST_CHECK
cv::CALIB_CB_FAST_CHECK
CV_RETR_CCOMP
cv::RETR_CCOMP
CV_CHAIN_APPROX_SIMPLE
cv::CHAIN_APPROX_SIMPLE
CV_CALIB_CB_FILTER_QUADS
cv::CALIB_CB_FILTER_QUADS
CV_GRAY2BGR
cv::COLOR_GRAY2BGR
CV_CALIB_CB_NORMALIZE_IMAGE
cv::CALIB_CB_NORMALIZE_IMAGE
CV_TERMCRIT_EPS
cv::TermCriteria::EPS
CV_TERMCRIT_ITER
cv::TermCriteria::COUNT
CV_THRESH_BINARY_INV
cv::THRESH_BINARY_INV
CV_CALIB_CB_FAST_CHECK
cv::CALIB_CB_FAST_CHECK
CV_CALIB_CB_ADAPTIVE_THRESH
cv::CALIB_CB_ADAPTIVE_THRESH
CV_THRESH_BINARY
cv::THRESH_BINARY
CV_SHAPE_CROSS
cv::MORPH_CROSS
CV_SHAPE_RECT
cv::MORPH_RECT
CV_ADAPTIVE_THRESH_MEAN_C
cv::ADAPTIVE_THRESH_MEAN_C
CV_FONT_HERSHEY_SIMPLEX
cv::FONT_HERSHEY_SIMPLEX
#include <opencv/cv.h>
#include <opencv2/opencv.hpp>
#include <opencv/highgui.h>
#include <opencv2/highgui/highgui.hpp>
到此为止已经完成VINS-Mono的安装
-
安装Realsense驱动
首先到Intel RealSense SDK 2.0官网下载,如果是Linux Ubuntu,找到https://github.com/IntelRealSense/librealsense/blob/master/doc/distribution_linux.md直接命令行安装
$ sudo apt-key adv --keyserver keyserver.ubuntu.com --recv-key F6E65AC044F831AC80A06380C8B3A55A6F3EFCDE || sudo apt-key adv --keyserver hkp://keyserver.ubuntu.com:80 --recv-key F6E65AC044F831AC80A06380C8B3A55A6F3EFCDE $ sudo add-apt-repository "deb https://librealsense.intel.com/Debian/apt-repo $(lsb_release -cs) main" -u $ sudo apt-get install librealsense2-dkms $ sudo apt-get install librealsense2-utils $ sudo apt-get install librealsense2-dev $ sudo apt-get install librealsense2-dbg
如果相机通过usb连接电脑,采用命令
realsense-viewer
可以查看图像安装ROS驱动
sudo apt-get install ros-noetic-realsense2-camera sudo apt-get install ros-noetic-realsense2-description mkdir -p ~/catkin_ws/src cd ~/catkin_ws/src/ git clone https://github.com/IntelRealSense/realsense-ros.git cd .. catkin_make source devel/setup.bash roslaunch realsense2_camera rs_camera.launch # 该命令正常打开相机
此时需要修改一下rs_camera.launch,IMU开启,相机分辨率640×480,重命名为
rs_camera_d435i.launch
<launch><arg name="serial_no" default=""/><arg name="usb_port_id" default=""/><arg name="device_type" default=""/><arg name="json_file_path" default=""/><arg name="camera" default="camera"/><arg name="tf_prefix" default="$(arg camera)"/><arg name="external_manager" default="false"/><arg name="manager" default="realsense2_camera_manager"/><arg name="fisheye_width" default="640"/><arg name="fisheye_height" default="480"/><arg name="enable_fisheye" default="false"/><arg name="depth_width" default="640"/><arg name="depth_height" default="480"/><arg name="enable_depth" default="false"/><arg name="infra_width" default="640"/><arg name="infra_height" default="480"/><arg name="enable_infra1" default="true"/><arg name="enable_infra2" default="true"/><arg name="color_width" default="640"/><arg name="color_height" default="480"/><arg name="enable_color" default="true"/><arg name="fisheye_fps" default="30"/><arg name="depth_fps" default="30"/><arg name="infra_fps" default="30"/><arg name="color_fps" default="30"/><arg name="gyro_fps" default="200"/><arg name="accel_fps" default="250"/><arg name="enable_gyro" default="true"/><arg name="enable_accel" default="true"/><arg name="enable_pointcloud" default="false"/><arg name="pointcloud_texture_stream" default="RS2_STREAM_COLOR"/><arg name="pointcloud_texture_index" default="0"/><arg name="enable_sync" default="true"/><arg name="align_depth" default="true"/><arg name="publish_tf" default="true"/><arg name="tf_publish_rate" default="0"/><arg name="filters" default=""/><arg name="clip_distance" default="-2"/><arg name="linear_accel_cov" default="0.01"/><arg name="initial_reset" default="false"/><arg name="unite_imu_method" default="linear_interpolation"/><arg name="topic_odom_in" default="odom_in"/><arg name="calib_odom_file" default=""/><arg name="publish_odom_tf" default="true"/><arg name="allow_no_texture_points" default="false"/><arg name="emitter_enable" default="false"/><group ns="$(arg camera)"><include file="$(find realsense2_camera)/launch/includes/nodelet.launch.xml"><arg name="tf_prefix" value="$(arg tf_prefix)"/><arg name="external_manager" value="$(arg external_manager)"/><arg name="manager" value="$(arg manager)"/><arg name="serial_no" value="$(arg serial_no)"/><arg name="usb_port_id" value="$(arg usb_port_id)"/><arg name="device_type" value="$(arg device_type)"/><arg name="json_file_path" value="$(arg json_file_path)"/><arg name="enable_pointcloud" value="$(arg enable_pointcloud)"/><arg name="pointcloud_texture_stream" value="$(arg pointcloud_texture_stream)"/><arg name="pointcloud_texture_index" value="$(arg pointcloud_texture_index)"/><arg name="enable_sync" value="$(arg enable_sync)"/><arg name="align_depth" value="$(arg align_depth)"/><arg name="fisheye_width" value="$(arg fisheye_width)"/><arg name="fisheye_height" value="$(arg fisheye_height)"/><arg name="enable_fisheye" value="$(arg enable_fisheye)"/><arg name="depth_width" value="$(arg depth_width)"/><arg name="depth_height" value="$(arg depth_height)"/><arg name="enable_depth" value="$(arg enable_depth)"/><arg name="color_width" value="$(arg color_width)"/><arg name="color_height" value="$(arg color_height)"/><arg name="enable_color" value="$(arg enable_color)"/><arg name="infra_width" value="$(arg infra_width)"/><arg name="infra_height" value="$(arg infra_height)"/><arg name="enable_infra1" value="$(arg enable_infra1)"/><arg name="enable_infra2" value="$(arg enable_infra2)"/><arg name="fisheye_fps" value="$(arg fisheye_fps)"/><arg name="depth_fps" value="$(arg depth_fps)"/><arg name="infra_fps" value="$(arg infra_fps)"/><arg name="color_fps" value="$(arg color_fps)"/><arg name="gyro_fps" value="$(arg gyro_fps)"/><arg name="accel_fps" value="$(arg accel_fps)"/><arg name="enable_gyro" value="$(arg enable_gyro)"/><arg name="enable_accel" value="$(arg enable_accel)"/><arg name="publish_tf" value="$(arg publish_tf)"/><arg name="tf_publish_rate" value="$(arg tf_publish_rate)"/><arg name="filters" value="$(arg filters)"/><arg name="clip_distance" value="$(arg clip_distance)"/><arg name="linear_accel_cov" value="$(arg linear_accel_cov)"/><arg name="initial_reset" value="$(arg initial_reset)"/><arg name="unite_imu_method" value="$(arg unite_imu_method)"/><arg name="topic_odom_in" value="$(arg topic_odom_in)"/><arg name="calib_odom_file" value="$(arg calib_odom_file)"/><arg name="publish_odom_tf" value="$(arg publish_odom_tf)"/><arg name="allow_no_texture_points" value="$(arg allow_no_texture_points)"/></include></group> </launch>
确认图像和IMU话题数据都有
rostopic echo /camera/imu rostopic echo /camera/color/image_raw
-
修改VINS-Mono的配置文件
在
VINS-Mono/config/realsense/realsense_color_config.yaml
中修改对应的imu和camera topic,输出路径,其中相机内参数,厂家出厂的d435i内参有少许不一致,需要修改一下。不想标定就直接用rostopic echo /camera/color/camera_info
,查看相机内参,不过最好自己标定一下,采用ROS提供的张正友标定法。imu与camera外參可以不管,imu白噪声和随机游走可以采用原始值,但最好也自己标定一下,参考网站。 -
运行
roslaunch realsense2_camera rs_camera_d435i.launch roslaunch vins_estimator vins_rviz.launch roslaunch vins_estimator realsense_color.launch
单目+IMU需要运动初始化,有时因IMU积分的原因而导致漂移严重。
D435i运行VINS-Fusion
这里采用D435i输出的红外双目图像,VINS-Fusion采用的是双目+IMU会进行估计,该方案不需要运动初始化,而VINS-Mono需要相机初始的时候运动初始化。安装基本和前面VINS-Mono的一致
cd ~/VINS-Fusion_ws/src
git clone https://github.com/HKUST-Aerial-Robotics/VINS-Fusion.git
cd ../
catkin build
source ~/VINS-Fusion_ws/devel/setup.bash
运行D435i时需要查看红外图像,同样是上述的rs_camera_d435i.launch
修改VINS-Fusionconfig/realsense_d435i/realsense_stereo_imu_config.yaml
其中estimate_extrinsic最好都设置成1,这样就会优化初值。
%YAML:1.0#common parameters
#support: 1 imu 1 cam; 1 imu 2 cam: 2 cam;
imu: 1
num_of_cam: 2 imu_topic: "/camera/imu"
image0_topic: "/camera/infra1/image_rect_raw"
image1_topic: "/camera/infra2/image_rect_raw"
output_path: "~/output/"cam0_calib: "left.yaml"
cam1_calib: "right.yaml"
image_width: 640
image_height: 480# Extrinsic parameter between IMU and Camera.
estimate_extrinsic: 1 # 0 Have an accurate extrinsic parameters. We will trust the following imu^R_cam, imu^T_cam, don't change it.# 1 Have an initial guess about extrinsic parameters. We will optimize around your initial guess.body_T_cam0: !!opencv-matrixrows: 4cols: 4dt: ddata: [ 1, 0, 0, -0.00552,0, 1, 0, 0.0051,0, 0, 1, 0.01174,0, 0, 0, 1 ]body_T_cam1: !!opencv-matrixrows: 4cols: 4dt: ddata: [ 1, 0, 0, 0.0446571,0, 1, 0, 0.0051,0, 0, 1, 0.01174,0, 0, 0, 1 ]#Multiple thread support
multiple_thread: 1#feature traker paprameters
max_cnt: 150 # max feature number in feature tracking
min_dist: 30 # min distance between two features
freq: 10 # frequence (Hz) of publish tracking result. At least 10Hz for good estimation. If set 0, the frequence will be same as raw image
F_threshold: 1.0 # ransac threshold (pixel)
show_track: 1 # publish tracking image as topic
flow_back: 1 # perform forward and backward optical flow to improve feature tracking accuracy#optimization parameters
max_solver_time: 0.04 # max solver itration time (ms), to guarantee real time
max_num_iterations: 8 # max solver itrations, to guarantee real time
keyframe_parallax: 10.0 # keyframe selection threshold (pixel)#imu parameters The more accurate parameters you provide, the better performance
acc_n: 0.1 # accelerometer measurement noise standard deviation. #0.2 0.04
gyr_n: 0.01 # gyroscope measurement noise standard deviation. #0.05 0.004
acc_w: 0.001 # accelerometer bias random work noise standard deviation. #0.002
gyr_w: 0.0001 # gyroscope bias random work noise standard deviation. #4.0e-5
g_norm: 9.805 # gravity magnitude#unsynchronization parameters
estimate_td: 1 # online estimate time offset between camera and imu
td: 0.00 # initial value of time offset. unit: s. readed image clock + td = real image clock (IMU clock)#loop closure parameters
load_previous_pose_graph: 0 # load and reuse previous pose graph; load from 'pose_graph_save_path'
pose_graph_save_path: "~/output/pose_graph/" # save and load path
save_image: 1 # save image in pose graph for visualization prupose; you can close this function by setting 0
出行效果如下:
但拿出去走廊走一圈回到原来的地方(但未检测到回环)效果如下:
完整版视频
VINS-Fusion with D435i
效果貌似还可以,如果后面检测到回环效果会更好。
总结
两种方案,个人认为双目+IMU(VINS-Fusion运行D435i的红外双目图像+IMU)比单目+IMU(VINS-Mono运行D435的RGB图像+IMU)更好,双目图像本身两张图像就可以三角化深度估计了,而单目图像需要运动时候三角化进行深度估计(有尺度漂移问题),所以会更好一些。