Comments (6)
Problem with this approach is that pose estimation requires target size to be specified. Options are:
- See if size can be encoded into targets
- Split recognition and estimation into multiple levels, do a first level 2d recognition and order by area/size, then attempt pose estimation in ascending order of size. Smallest target with successful pose estimation wins, as this will be the physically smallest (and hence most accurate and reliable) target.
from vision_landing.
Add a capture threshold to debounce transition from one target to another.
from vision_landing.
Refactored code to do ordered pose estimation by marker area (size) order ascending, so smallest marker first. A static size mapping added, so:
// Temp declaration of marker sizes
map<uint32_t,float> markerSizes = { {12,0.5}, {36,0.24}, {85,0.11}, {161,0.11}, {166,0.11}, {227,0.11} };
If a size mapping exists for the marker id it uses that to perform pose estimation, else it uses the generic markersize config value, but it chooses the smallest marker with a valid size mapping as the active marker.
Chosen marker bounces a lot during transition, needs a threshold/debounce applied so marker transitions only happen when smaller marker is consistently captured across multiple frames.
Also need to add markerSizes mapping as a config variable and to somehow decode into c++ map.
from vision_landing.
Added --sizemapping (or -z) config setting that allows setting marker_id:size mappings. This allows the use of multiple markers with different sizes.
sizemapping=12:0.5,36:0.24,85:0.11,161:0.11,166:0.11,227:0.11
This is after parsing within track_targets:
Info['Size Mappings', '12=0.5, 36=0.24, 85=0.11, 161=0.11, 166=0.11, 227=0.11,']
from vision_landing.
Working nicely now, still need to add a debounce/threshold mechanism to stop fast transitions between markers that might upset flight control.
from vision_landing.
First attempt at adding a mechanism to stabilise the transition between markers. It tracks the history of each marker over the past 10 frames, and will only activate a new marker if the marker history shows it has been detected and estimated for >50% of the frame history. It helps but needs refinement, in particular the frame history size and threshold need to be configurable parameters, tracked in a separate issue.
from vision_landing.
Related Issues (20)
- Add mavlink target fields to TIMESYNC and LANDING_TARGET
- Could you give 1 more line of description in your warning? HOT 2
- Take timestamp as close to shutter as possible HOT 4
- track_targets takes long time to shut down HOT 2
- Python ThreadedReader queue building up HOT 1
- Add config param for PLND_EST_TYPE
- Aruco camera calibration not working
- Vision Landing on raspberry Pi3 B+ HOT 2
- Migrate to python 3
- Update vision_landing.conf sample config HOT 1
- Create better default values for threshold/history HOT 1
- global name 'timespec' is not defined HOT 3
- aruco make / install instructions incorrect HOT 3
- Migrate to apriltag HOT 3
- Nothing happens when I run sudo systemctl start vision_landing HOT 1
- Error while running cmake . && make && make install
- Can this run on jevois ?
- Why deprecated? HOT 9
- Define a landing point relative to all markers HOT 2
- Licensing change
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from vision_landing.