说明：使用视觉-惯性和纯激光里程计，通过因子图后处理全局位姿。其中平移的权重激光比例高，旋转的权重视觉-惯性比例高 其中视觉前端特征点使用激光点云和三角化共同估计深度（添加mask-rcnn剔除动态特征点）, 激光前端采用帧到局部地图的特征最近邻检索匹配，后端优化计算雅克比矩阵，迭代优化，保证构建地图的精度 传感器类型：大恒工业相机、Xsens惯性传感器、VLP-32激光雷达、松灵机器人底盘

This work is based on f-loam and VINS-Mono, On this basis, We use the laser point to obtain the depth and factor map and optimize the way to realize the pose estimation.

This code is modified from F-LOAM and VINS-Mono and SC-A-LOAM.

• visual_inertial
• lidar_fsloam
• global_fusion
• ScanContext loop detect

• [Ubuntu-16.04]
• [ROS-kinetic]
• [Ceres-2.0.0]
• [OpenCV-3.3.1]
• [PCL-1.7.2]
• [gtsam-4.0.2]
• [Eigen-3.3.7]
• [OpenMP]

    cd ~/catkin_ws/src
    git clone https://github.com/GuoFeng-X/Sensor_Fusion.git
    cd ..
    catkin_make -j4

    cd catkin_ws
    source devel/setup.bash
    roslaunch Sensor_fusion module_visual.launch

    cd catkin_ws
    source devel/setup.bash
    roslaunch Sensor_fusion module_fsloam.launch

    rosbag plag kitti_07.bag --pause --clock -r0.5

Using database KITTI-07 run this code.

下载基于ROS话题发布的mask-rcnn前端代码。mask-rcnn ros版本

在动态场景ADVIO数据集下测试

Copy the above code to the Sensor_fusion folder

    cd catkin_ws/src/Sensor_fusion
    unzip Mask-RCNN
    cd script/mask_rcnn
    ./run_build.sh
    cd detectron2/Visual_Frand
    ./run_detect_mask.sh

    cd catkin_ws
    source devel/setup.bash
    roslaunch Sensor_fusion module_visual_mask.launch

    rosbag plag ADVIO-05.bag --pause -r0.5

Thanks for F-LOAM and VINS-Mono.

• F-LOAM(Wang H, Wang C, Chen C L, et al. F-loam: Fast lidar odometry and mapping.)
• VINS-Mono(Tong Q, Peiliang L, et al. VINS-Mono: A Robust and Versatile Monocular Visual-Inertial State Estimator.)