perception-multi-sensor-fusion

Introduction

The multi-sensor fusion module fuses the output results of Lidar, Camera, and Radar multiple sensors to make the detection results more reliable.

It uses post-processing technology, the algorithm used is probabilistic fusion.

Directory Structure

multi_sensor_fusion
├── BUILD      // bazel build file
├── README.md
├── base       // base data structure
├── common     // filter and utils
├── conf       // dag config file
├── cyberfile.xml  // package description file
├── dag
├── data        // lib's configuration file
├── fusion      // fusion function
├── interface
├── multi_sensor_fusion_component.cc  // component
├── multi_sensor_fusion_component.h
└── proto       // proto file

Input and Output

MultiSensorFusionComponent

apollo::perception::fusion::MultiSensorFusionComponent

Input

Note: The input channel is structure type data. The default trigger channel is /perception/inner/PrefusedObjects. The detailed input channel information is in modules/perception/multi_sensor_fusion/dag/multi_sensor_fusion.dag file. By default, the upstream components of the messages received by the component include lidar_detection_tracking, camera_tracking, radar_detection.

Output

Note: The output channel is proto type data. The detailed output channel information is in modules/perception/multi_sensor_fusion/conf/multi_sensor_fusion_config.pb.txt file.

How to run

The multi-sensor fusion module does not support running alone, it needs to run together with lidar, camera and radar detection modules.

You can use the following command to start the whole perception function, including lidar, camera, and radar target detection, and output their results after fusion.

cyber_launch start modules/perception/launch/perception_all.launch