Visual_occupancy_map

A ROS node for generating occupancy grid map and semantic occupancy grid map based on visual SLAM.

Dependencies

Tested on Ubuntu 18.04, ROS Melodic, OpenCV 3.2.0, Eigen 3.2.10.

• OpenCV;
• Eigen3.

Build

You can use provided Dockerfile or build using catkin build.

Subscribed topics

• /KF_pose - nav_msgs::Odometry - pose of camera in a keyframe;
• /KF_map_points - sensor_msgs::PointCloud - map points that are visible on a keyframe;
• /semantic_image - sensor_msgs::Image - image with result of semantic segmentation (only for semantic_occupancy_map_node);

Published topics

• /visual_occupancy_node/grid_map - nav_msgs::OccupancyGrid - occupancy grid map;
• /visual_occupancy_node/trajectory_grid_map - sensor_msgs::Image - image with occupancy grid map and trajectory of robot;
• /visual_occupancy_node/semantic_occupancy_map - sensor_msgs::Image - semantic occupancy map (only for semantic_occupancy_map_node);

Parameters

You can change next parameters in .yaml file:

• Scale_factor - 1 / resolution of map - default: 20;
• Tcw_form - form of KF_pose, 0 equals Twc form - default: 0;
• Cloud_max_x - max value of map's horizontal in meters - default: 10;
• Cloud_min_x - min value of map's horizontal in meters - default: -10;
• Cloud_max_z - max value of map's vertical in meters - default: 16;
• Cloud_min_z - min value of map's vertical in meters - default: -5;
• Free_thresh - value above which cell determines as free area - default: 0.55;
• Occupied_thresh - value below which cell determines as occupied area - default: 0.5;
• Use_local_counters - using local counters in keyframes for reduce fake free cells - default: 0;
• Use_gaussian_counters - using gaussian smoothing for reduce discretization - default: 0;
• Use_boundary_detection - using Canny algorithm for detect boundary between free and unknown area and set it as occupancy area - default: 0;
• Publish_trajectory - publish image in '/visual_occupancy_node/trajectory_grid_map' topic - default: 0;
• Visit_thresh - visit counter value above which map points are provided for update occupancy map - default: 0;
• Thresh_extension - when difference between KF_pose and bound of map in meters below this value, extension of map is starting - default: 5;
• Step_extension - value which added to bound of map when extension is running - default: 10;
• Gaussian_kernel_size - kernel size for gaussian smoothing - default: 3;
• Canny_thresh - value of lower thresh in Canny detector, upper thresh equals 2 * lower thresh - default: 350;
• Upper_height - map points which have height value above this thresh not used - default: 1.0;
• Lower_height - map points which have height value below this thresh not used - default: 0.0;
• Make_submaps - make submaps whith colours in submaps.txt (only for semantic_occupancy_map_node) - default: 0;
• Use_semantic_filter - using only map points whith semantic labels in submaps.txt for update counters (only for semantic_occupancy_map_node) - default: 0.

Examples

For running occupancy_map_node with provided config file sber_mseg.yaml use:

roslaunch visual_occupancy_map sber_mseg.launch

For running semantic_occupancy_map_node with provided config file sber_mseg.yaml use:

roslaunch visual_occupancy_map semantic_sber_mseg.launch

If you want to build additional submaps, setup submaps.txt and use:

roslaunch visual_occupancy_map submaps.launch

Also you can use configs occupancy_map_node and semantic_occupancy_map_node for rviz in folder 'rviz'.