Orchard Mapping
Accurate mapping will be the foundation of any system of autonomous robotics in orchards. For a robot to navigate it first needs to know where it is with high accuracy in real time. It will also be essential for building the geo-referenced information systems that will allow high level presentations of data and will support better decisions on farm.
However this is very difficult to achieve in tall wooded environments. All flavours of GPS right through to RTK suffer from loss of accuracy when they don't have a clear view of the horizon - in orchards it is not unusual to see a DGPS system rated for 1m accuracy drift more than 10m and sometimes the signal is lost completely. Manual surveying is not an option as it is far too expensive.
In short there is a real need for an orchard positioning system that is
I have been working on this problem for a long time really, as have many others I imagine. Also in common with many others this project uses data streamed from a variety of sources - GPS, Inertial Measurement, Video - and attempts to fuse this data into a accurate real time orchard position with a level of robustness such that it can handle intermittent streams from the data sources when necessary.
In general the project uses low cost hardware and sensors with a goal of a simple plug-in system that can be used with a variety of equipment. It does not use SLAM (simultaneous location and mapping). Rather it requires an initial drive-through of the block after which the system self-updates, this results in an intrinsically structured map that can make it easier to integrate with legacy data.
Again like everyone else success is just a couple of trials away :-)
The map below was not made with the system described above, rather it was produced by manually stepping through the process that is currently under development. As such it tests several aspects of the envisaged algorithms, data handling and data visualisation.
In the map tree yields are represented by the colour - red being poor and blue good; and the tree canopy volume by the size of the boxes.
Looking carefully at the map it is clear there are still some location errors to be fixed. It is also interesting to note the level of variability in yield and that there are both very large and very small structures in the variability, indicating that Precision Agriculture techniques applied down to the single tree level will have significant benefits for tree crops.
To see the whole thing it needs to be viewed in Google Earth with this file. Testkml.kml
Back To Home Copyright © 2016 David Bell, Australia. All
Rights Reserved.
No parts of this web site may be copied without the
permission of the author.