GhiniTreePositioner

what is this

This plugin solves the problem of estimating accurate positions for trees under a thick canopy, or for locations where high resolution aereal photos are not available.

how does it work

Points need a 'id' property, and mutual distances must be measured in metres. Mutual distances are given in the form of a comma separated file, first two columns are the 'id' of to points, third column their distance.

The plugin implements two separate functionalities:

1. new points

Points already in the layer will be used as reference, Points referred to in the mutual distances file, but not in the layer, they will be added to the layer, as precisely as possible

2. GPS correction

You work on two layers, one with GPS data, one with the output from the plugin. The mutual distances network will be used to compute the correct pattern on the ground, the precise relative position of the points. This need not necessarily match the GPS data. This pattern is then rigidly moved to minimize the square distances to the corresponding GPS locations.

does it really work

I consider this plugin still at a rather experimental stage, but "yes, of course it works!" I will be enhancing it as I need that, or as reaction to user feedback.

It would be highly useful if you could provide me with example data, for which, I'm quite sure, the plugin does not yet work.

show me an example

we are in Colombia, a small area near La Macarena, I have a GeoTiff for the area and we're looking at it, and 4 reference points, they have an 'id' field and they are called, clockwise from bottom left, corner, source-2, source, and junction:

we do not own a GPS machine, or maybe the battery was down, and we observed two trees in the middle of this area, and we could measure, with the approximation of 0.5m, the distances of tree A to source-2, source, corner and junction, and of tree B to source, junction, and the other tree.

we put this information in a csv file, like this:

A,source2,70
A,corner,79,
A,source,157
A,junction,154
B,A,58
B,source,148.5
B,junction,98.5

then we invoke the plugin, specifying the name of the layer, and the csv file holding the distances

and we get the result in the same layer.

what about the GPS correction

this is slightly more sofisticated. imagine you have GPS point measurements which you quite rightfully do not blindly trust. so what you do is you measure mutual distances among physical points, and you do trust the correctness of these mutual distances.

this GPS correction tool uses the mutual distances to generate a rigid frame which it fits with the GPS point measurements. the resulting points set minimizes the sum of square distances from the points as coming from your GPS device.

graphically, we have the following set of points. this represents the real situation on the ground, which GPS measurements can only approximate.

this is the GPS approximation of reality, plotted together with reality, which we hope to approximate better. as you can appreciate, the shape of the patterns formed by the GPS approximation are so different from the real patterns that a map based on only these GPS points would be difficult to use in practice.

this picture shows the data we feed to the GPS correction tool: the GPS data, and a numerical matrix of mutual distances, which we here show as a rigid frame. don't be confused by the position of the frame, we don't really know where to put the frame.

the result is a better approximation of the real situation on the ground, as you can see here, where we first compare it to the GPS measurements and to the unknowable reality.

the result of our GPS correction tool respects the provided mutual distances, and uses the full set of GPS points to better approximate reality.