The QGIS Green View Index plugin is a powerful tool to measure the vegetation cover of a given area. It is interesting to see what it brings compared to the classical processing based on vertical aerial photos.
We publish a series of three articles:
1-Concepts of the Green View index and installation of the plugin
2-Tutorial of the Green View Index plugin.
3-Comparison of the results of the plugin and the aerial photo processing (this article)
We are going to compare the green index calculated on two aerial photos of the city center of Brest with the processing of the Green View Index plugin of QGis.
For this we downloaded two aerial images from the Brest Métropole website.
Each of them covers an area of about 1 km².
Generating Sample Points
For the calculation of the Green View Index, it is necessary to define an area of interest (AOI). We will therefore create a new layer of polygon type corresponding to each aerial image.
Use the menu Layer-> Create a layer -> New Shapefile layer
For the second photo we repeat the operation.
So we have our two AOIs. We can run the first script of the Green View Index plugin, Generate Sample Points, to create a layer of points randomly distributed on each of our AOIs.
1-Using AOI only
We use the default settings, i.e. creating 100 points distributed over the AOI. We did not use a road network.
We run the script on each of the AOIs and we obtain two layers of type point: pointsSouth and pointsNorth.
We can then execute the second script of the plugin,Download Google Street View Images.
We have kept the default settings here too: the six directions around the viewpoint and the three view tilts.
In the Log tab you can see the number and the points for which there was no image available.
For the North zone, 5 points have no images. For the South zone 19 points do not give any result.
While browsing the image directories, we noticed a rather peculiar phenomenon. Indeed, some viewpoints located on buildings return images… of the interior of the buildings!
Another “anomaly” that we noticed is the presence of shots in the middle of the night.
It is obvious that these images must be excluded from the calculation. In summary we obtained
NORTH Zone: Out of 100 viewpoints, 5 points without images, 4 points inside buildings and 1 night viewpoint, that is 10% of non-usable points.
South Zone: Out of 100 viewpoints, 19 points without images and 2 points inside buildings, that is 21% of non-usable points.
2- Using AOI and a road network.
We therefore downloaded the Brest road network to test the joint use of the AOIs and the network.
The following image shows the position of the 100 points resulting from the use of AOIs and the road network (⋇) as well as the position of the 100 points resulting from the use of only the AOI (∆).
The same checks of the images downloaded from Google Street View were carried out and obtained:
NORTH Zone: Out of 100 view points, 3 points without images, no points inside buildings and no night view points, that is 3% of points not usable.
South Zone: Out of 100 viewpoints, no points without images and 1 point inside buildings, that is to say 1% of non-usable points.
It seems clear that, in dense urban areas, the use of the road network is highly necessary. The difference in results will be seen later.
Green View Index calculation
The Green View Index is calculated for each area and each type of point generation
The results of the average GVI obtained are as follows:
North Zone
All AOI points……………………………… 0.061
AOI points without anomalies…….. 0.064
All AOI and roadway points…………. 0.051
South Zone
All AOI points……………………………… 0.034
AOI points without anomalies……. 0.034
All AOI and road points……………….. 0.047
This small project is, of course, not sufficient to statistically determine the best results, but it shows that the results depend much more on the type of point generation than on the presence of anomalies in the Google Street View images.
Comparison with aerial images.
We used the plugin to process both aerial photos with the same algorithm:
The images show the original image and the image with the pixels determined as “green”.
The GVI of the North image is calculated as 0.1, and that of the South image as 0.07.
Both are significantly higher than the average GVI calculated on the Google Street View images: 0.1 versus 0.05 and 0.07 versus 0.047 respectively.
Conclusions
This small test does not claim to determine whether the Green View Index plugin gives “better” results than the traditional method of calculating from aerial photos. It simply shows that the results are different. The task now is to understand why and what more they can bring to the knowledge of the urban environment.
