i/rtabmap
1
0
Fork 0
mirror of https://github.com/introlab/rtabmap.git synced 2026-10-09 11:37:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

Updated Loop closure detection (markdown)

matlabbe
2016-12-05 16:19:39 -05:00
parent 295ae4f684
commit 6a61d86ef5
1 changed files with 3 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files
Loop-closure-detection.md
+3 -17
@@ -15,12 +15,6 @@ This simple example shows how to use RTAB-Map to process a data set.
[[doc/Tutorials/ProcessADataSet/main_window.png]]
* Clean the working directory. Menu Edit -> "Delete memory (# Mb)". A warning is shown about erasing the database. If you don't need to keep it, click ok. Otherwise, click no, then Menu Edit -> "Show working directory in file browser" and backup the database (LTM.db) and log files (LogI.txt, LogF.txt) before executing the "Delete memory" action.
[[doc/Tutorials/ProcessADataSet/menu_delete_memory.png]]
[[doc/Tutorials/ProcessADataSet/dialog_delete_mem.png]]
* Setup parameters for appearance-based mapping only. Download this file [settingsAppearanceDemo.ini](https://raw.githubusercontent.com/wiki/introlab/rtabmap/doc/Tutorials/ProcessADataSet/settingsAppearanceDemo.ini), open Edit->Preferences->General settings. Click on "Reset settings" to start from default parameters, then click on "Load settings" and select the file you have just downloaded. Here are information about which parameters are modified from the default ones:
* RGBD/Enabled = False (disable metric stuff)
* Rtabmap/ImageBufferSize = 0 (process all images)
@@ -31,17 +25,13 @@ This simple example shows how to use RTAB-Map to process a data set.
* Mem/BadSignaturesIgnored = true (ignore images with little or no features)
* Setup the source of the camera. For this example, download this [sample data set](https://raw.githubusercontent.com/introlab/rtabmap/master/bin/data/samples.zip) and extract it anywhere. Open menu Detection -> "Select Source" -> "More options..." (or Preferences->Source panel), then select "Source" -> "RGB". Scroll down to RGB section and select "Images" instead of "Usb camera", then select the directory containing images just extracted. Make sure the input source is not 0 Hz (otherwise the images will be processed very fast).
[[doc/Tutorials/ProcessADataSet/menu_source_images.png]]
[[doc/Tutorials/ProcessADataSet/menu_source.png]]
* Click on "Start" button in the toolbox of the main window. You should see briefly an initialization dialog and the detection is paused on the first image. The initialization can be done before starting (Menu Edit -> "Apply settings to the detector"), thus the detection will not be paused when starting the detection.
[[doc/Tutorials/ProcessADataSet/dialog_initialization.png]]
* Click on "New database", then "Start" button in the toolbox of the main window. You should see briefly an initialization dialog and the detection will start;
[[doc/Tutorials/ProcessADataSet/main_detection_start.png]]
* Push button "Pause on match". This will make the detector pause when an accepted loop closure is found.
* Push button "Pause" to continue the detection. The detection should work until the image 50 is acquired.
* You can click on button "Pause on match" to make the detector pause when an accepted loop closure is found.
[[doc/Tutorials/ProcessADataSet/main_loop_detected.png]]
@@ -52,10 +42,6 @@ This simple example shows how to use RTAB-Map to process a data set.
* Yellow = same as Green, but found more than one time in the image.
* Pink = feature found in the two images.
* Push button "Pause on match" to avoid pausing on each loop closure detected.
* Push button "Pause" to continue the detection to the end of the data set.
* When the last image (84) is processed, click Menu Window -> "Show view" -> Statistics. This will open a panel showing statistics. For this tutorial, look only information in the top part of the panel. We can see the number of loop closures detected/rejected.
[[doc/Tutorials/ProcessADataSet/main_statistics.png]]