Module Name

lidar_segmentation

Introduction

The lidar segmentation module uses point cloud of secondary_indices to do object segmentation. Secondary indices is intersetion of non_ground_indices and roi_indices, and remove foreground objects point indices.

Directory Structure

├── lidar_segmentation     // lidar segmentation component
    ├── common             // common codes
    ├── conf               // configuration folder
    ├── dag                // module startup file
    ├── data               // module configuration parameters
    ├── launch             // launch file
    ├── proto              // lidar segmentation module configuration proto file
    ├── segmentor          // point cloud segmentation methods
    ├── lidar_segmentation_component.cc // component entrance
    ├── lidar_segmentation_component.h
    ├── cyberfile.xml      // package management file
    ├── README.md          // readme file
    ├── README_cn.md
    └── BUILD              // compile file

Modules

LidarSegmentationComponent

apollo::perception::lidar::LidarSegmentationComponent

Input

Name	Type	Description
`msg`	`onboard::LidarFrameMessage`	lidar frame message

Output

Name	Type	Description
`frame`	`onboard::LidarFrameMessage`	lidar frame message

Configuration

Configurations of graph_segmentation.

parameter type	parameter name	default value	meaning
uint32	grid_width		grid width
uint32	grid_height		grid height
float	resolution		grid resolution
float	threshold		graph segmentation threshold
uint32	min_pt_number		min point number of cluster
uint32	search_radius	3	grid search radius
float	height_threshold	2.0	height threshold decided by autonomous car
float	xmin	-30.0	pointcloud valid distance of front
float	xmax	30.0	pointcloud valid distance of back
float	ymin	-10.0	pointcloud valid distance of right
float	ymax	30.0	pointcloud valid distance of left
float	semantic_cost	1.0	semantic label cost
float	same_semantic_coefficient	1.0	edge coefficient for same semantic label
float	diff_semantic_coefficient	1.0	edge coefficient for different semantic label
float	same_motion_coefficient	1.0	edge coefficient for same motion label
float	diff_motion_coefficient	1.0	edge coefficient for different motion label
float	split_aspect_ratio	10.0	split aspect ratio
float	split_distance	5.0	split distance

How to Launch

Add vehicle parameter configuration file to modules/perception/data/params in profile, corresponding frame_id and sensor_name, launch transform

bash

cyber_launch start /apollo/modules/transform/launch/static_transform.launch

Modify modules/perception/launch/perception_lidar.launch

select the dag file to start, add lidar_segmentation.dag to perception_lidar.launch
modify msg_adapter. It is used to wrap messages sent by other steps as /apollo/perception/obstacles, this can be used for individual debugging. Modify relevant channel configurations in modules/perception/data/flag/perception_common.flag

Modify parameters of modules/perception/lidar_segmentation/conf/lidar_segmentation_config.pb.txt in profile

output_channel_name: output channel name
plugin_param: plugin parameters
- name: method name
- config_path: configuration folder
- config_file: configuration file name

Launch pointcloud perception

bash

cyber_launch start /apollo/modules/perception/launch/perception_lidar.launch