pcl-segment-lettuce

pcl-segment-lettuce

pcl-segment-lettuce segments lettuce-of-interest from coloured 3D point clouds. Below, the most common usage of pcl-segment-lettuce is presented.

Usage

The following sections show examples of how to use pcl-segment-lettuce. The examples are based on the provided examples and assume, that a terminal window is opened in one of the data example folder.

Help

Display help and input arguments:

pcl-segment-lettuce --help

Show segmentation steps

Show the point cloud and the segmentation steps in a interactive window, before closing the program.

pcl-segment-lettuce --input-file pc-scaled.pcd --plant-csv center.csv --bed-neighbor-locations pc-neighbor-centers.pcd --show-clouds 1

Use page-up, page-down, home and end to skip between the steps in the segmentation process. A short description of each step/point cloud is displayed in the window title. For more information about the functionality of the visualizer, press h while it is open.

Segment and save lettuce-of-interest

Segment lettuce-of-interest from a point cloud pc-scaled.pcd and save the segmented cloud pc-plant.pcd.

pcl-segment-lettuce --input-file pc-scaled.pcd --plant-csv center.csv --bed-neighbor-locations pc-neighbor-centers.pcd --output-file pc-plant.pcd

Segment and compare to ground truth labels

Segment a point cloud and compare it to a labelled point cloud. Providing a labelled point cloud, outputs a confusion matrix as well as precision, recall and F1-score for the provided point cloud.

pcl-segment-lettuce --input-file pc-scaled.pcd --plant-csv center.csv --bed-neighbor-locations pc-neighbor-centers.pcd --labels labels.pcd

Example output:

Confusion matrix:
         :    Lettuce      Other   :     Total 
.........:. .......... .......... .:............
Other    :        759      95741   :     96500 
Lettuce  :      99359      22549   :    121908 
Neighbor :       3444     128712   :    132156 
.........:. .......... .......... .:............
Total    :     103562     247002   :    350564 

Precision: 0.959416
Recall   : 0.815033
F1-score : 0.881350

Segment using optimized settings

To segment the provided examples using the same settings used in the paper, use the following commands for Cos and Iceberg, respectively. Replace <other-input-arguments> with any of the input argument examples above.

Cos:

pcl-segment-lettuce <other-input-arguments> --plant-height-th 4.8723 --leaf-smooth-th 4.1861 --leaf-inner-dist-th 0.15214 --leaf-outer-dist-th 0.18838 --leaf-min-size 542

Iceberg:

pcl-segment-lettuce <other-input-arguments> --plant-height-th 3.8647 --leaf-smooth-th 4.1719 --leaf-inner-dist-th 0.15304 --leaf-outer-dist-th 0.16766 --leaf-min-size 717