roi_cluster_fusion#
Purpose#
The roi_cluster_fusion is a package for filtering clusters that are less likely to be objects and overwriting labels of clusters with that of Region Of Interests (ROIs) by a 2D object detector.
Inner-workings / Algorithms#
The clusters are projected onto image planes, and then if the ROIs of clusters and ROIs by a detector are overlapped, the labels of clusters are overwritten with that of ROIs by detector. Intersection over Union (IoU) is used to determine if there are overlaps between them.

Inputs / Outputs#
Input#
| Name | Type | Description |
|---|---|---|
input |
tier4_perception_msgs::msg::DetectedObjectsWithFeature |
clustered pointcloud |
input/camera_info[0-7] |
sensor_msgs::msg::CameraInfo |
camera information to project 3d points onto image planes |
input/rois[0-7] |
tier4_perception_msgs::msg::DetectedObjectsWithFeature |
ROIs from each image |
input/image_raw[0-7] |
sensor_msgs::msg::Image |
images for visualization |
Output#
| Name | Type | Description |
|---|---|---|
output |
tier4_perception_msgs::msg::DetectedObjectsWithFeature |
labeled cluster pointcloud |
~/debug/image_raw[0-7] |
sensor_msgs::msg::Image |
images for visualization |
Parameters#
Core Parameters#
| Name | Type | Description |
|---|---|---|
use_iou_x |
bool | calculate IoU only along x-axis |
use_iou_y |
bool | calculate IoU only along y-axis |
use_iou |
bool | calculate IoU both along x-axis and y-axis |
use_cluster_semantic_type |
bool | if false, the labels of clusters are overwritten by UNKNOWN before fusion |
only_allow_inside_cluster |
bool | if true, the only clusters contained inside RoIs by a detector |
roi_scale_factor |
float | the scale factor for offset of detector RoIs if only_allow_inside_cluster=true |
iou_threshold |
float | the IoU threshold to overwrite a label of clusters with a label of roi |
unknown_iou_threshold |
float | the IoU threshold to fuse cluster with unknown label of roi |
rois_number |
int | the number of input rois |
debug_mode |
bool | If true, subscribe and publish images for visualization. |