Compatibilty with Melodic, remove opencv3 dependancy
opencv3 is no longer a package in melodic, so we use cv_bridge
to get the correct opencv transitively
New fusion with planar-based error estimation
* Estimate variance of fiducial obsevations based on how well they place the robot on the floor upright
* use Kalman + David method of fusing estimates
* split out Transform with Variance, do some cleanup
* Simple unit tests for core fusion code
* remove hokey outlier rejection
Add publish_tf rosparam functionality and updated launch file accordingly (#154)
* Added publish_tf rosparam to disable publishing pose tf
* Change default publish_tf to True and fix 'if' formatting
Contributors: Jack Kilian, Jim Vaughan, Rohan Agrawal
0.10.0 (2018-10-13)
use sum in quadrature by default
Disable multifiducial estimate by default
Added a tool for fitting a plane
Add map_read_only param, publish camera_pose
Contributors: Jim Vaughan, Rohan Agrawal
0.9.0 (2018-09-12)
Merge pull request #112 from UbiquityRobotics/davecrawley-patch-1
Added in a systematic error parameter
fix missed rename alexy to sum in quadrature
Suggested changes
Updating the launch file to match the changes to map.cpp
Added in a systematic error parameter
Added in a systematic error parameter. This parameter is intended to take in to account systematic (non-random) errors in the measurement of fiducial position. Non-random errors can be very pernicious, and can generate highly unexpected and difficult to resolve issues. For example if all the fiducials are slightly the wrong size this will not show up as a random variance. Similar systematic errors include, camera mis-calibration, sampling defects, digitization errors, lens flaring and so forth. Measurement accuracy in-homogeneity can also be considered a systematic error and this will also help with that issue although it probably should be treated explicitly rather than through this "hack". Ultimately the correct approach is to rigorously compute all these systematic errors and add them in to the model - we are simply saying that we think the systematic errors are less than the specified value.
Contributors: David Crawley, Rohan Agrawal
0.8.4 (2018-08-26)
Don't publish pose if camera position is not known
Update README.md
Contributors: Jim Vaughan, Rohan Agrawal
0.8.3 (2018-02-26)
Keep publishing map -> odom tf even if not fiducials are visible (#103)
Contributors: Jim Vaughan
0.8.2 (2018-02-14)
Add service call to clear the map
Contributors: Jim Vaughan
0.8.1 (2018-01-21)
Fixed fiducial_pose topic's frame_id from camera to map.
System Message: WARNING/2 (<string>, line 61)
Block quote ends without a blank line; unexpected unindent.
Contributors: Rohan Agrawal, nav-go
0.8.0 (2018-01-14)
0.7.5 (2017-12-06)
0.7.4 (2017-11-09)
Fix typos
Log full 6DOF pose for ALL and MUL
Pass previous rvec and tvec to solvePnP()
Added (multi) pose estimation to fiducial_slam (disabled by default)
Contributors: Jim Vaughan
0.7.3 (2017-07-16)
Install launch files and fiducials.rviz
add test of auto init
Print out 6DOF camera pose
Fix multiplication order bug in autoInit()
Renamed some variables to be more clear
Add publish_6dof_pose param to disable squashing of estimated robot pose
Contributors: Jim Vaughan
0.7.2 (2017-05-24)
Using std::isnan() to stop Debian Jessie build errors
Contributors: Jim Vaughan
0.7.1 (2017-05-22)
0.7.0 (2017-05-21)
Fix dependencies
Make sure that the variance sent to rviz doesn't truncate to 0
Rewrite the code in C++, kill the python
Fiducial transforms are always camera->fiducial
Moved all service and message definitions to fiducial_msgs
Contributors: Jim Vaughan, Rohan Agrawal
0.6.1 (2017-02-06)
0.6.0 (2017-02-04)
Split Python into separate files
Ddded median filter option
Better exception/handling
Contributors: Jim Vaughan, Rohan Agrawal
0.5.1 (2016-12-28)
Map pub srv (#36)
* Added publishing of map (#28) and reset service call (#35)
* Updated documentation