Estimation of vegetated surfaces with Computer Vision: how we improved and scaled up our model

Here at nam.R, we’re putting our hard work into building a Digital Twin of the French territory. We aggregate, clean and re-organize a large quantity of data of different formats coming from many different providers.

How we improved our estimation of vegetated surfaces with Computer Vision

Here at nam.R, we’re putting our hard work into building a Digital Twin of the French territory. We aggregate, clean and re-organize a large quantity of data of different formats coming from many different providers. Among them are aerial images (geo-referenced photographs taken by an airplane), which play a crucial role and represent one of the richest sources of information of the area they describe. You’ll discover with this article how at nam.R we have improved our estimation of vegetated surfaces.

Today, we want to take you through the study of a very meaningful case for us, the detection of vegetated surfaces around buildings. In this project, we will show you how we build a dataset suited to our application and what we went through. The detection of vegetation is a simple yet powerful tool to help understand the quality of life around the populated areas.

We decided to focus on the vegetated zones’ segmentation, which means we predict a class between “vegetation” or “not vegetation” for each pixel.

Simple ideas don’t always lead to good results:

Our first idea for detecting vegetated areas consisted on a very simple thresholding of the HSV value:

nam.R image HSV wheel maison computer vision
HSV wheel and hues selected for vegetation detection

We could consider as vegetation pixels with a hue between 40 and 160 for example (values we chose empirically). This encompasses the yellowish greens as well as more blue-green hues. We’d need to take into account saturation and value (luminance) as well in order to consider the scene luminosity and the image’s saturation. This was our first model, which consisted basically a filter on 3 dimensions. In order to smooth out the pixel classification we run standard morphological operators, closing and opening, to help regularize the detections spatially.


estimation of vegetated surfaced nam.R
Test image, filtered hue, smoothed out mask, image and detection overlap


This is a very cheap task that could make it easy to process the whole territory. However, this cheapness also makes it very unreliable. Indeed, shadows were poorly detected and the pixels classified as vegetation using such an algorithm could as well be part of a green colored roof, green shades of water or even simple chromatic aberrations in our data!


computer vision nam.R
Vegetation detection through the hue value doesn’t always work



Discover the full article here


More articles

  • Partnering for Progress: How Real-time Social Data Advances the SDGs

    SDGs: Sustainable Development Goals On September 23rd, during the the International UN Summit on Climate taking place in New York, UN Global Pulse, Dataminr and Twitter planed an evening of discussions and presentations of concretes examples of technologies and innovations […]

  • nam.R at European AI for Finance

    “Time to be concrete”  nam.R was at Ai for finance, a unique opportunity of networking at the european scale for all the actors of the AI in finance, the fintech meeting of the reopening season had as topic: “Time to […]

  • nam.R was at Impact AI

    nam.R at Impact AI On July 8th, nam.R was at Impact AI for their first annual conference. Laurence Lafont – as President of Impact AI, announced that the first white paper of Impact AI : “A collective commitment for a […]

  • A week at Data Science Summer School 

    A week at DS3 After the success of the 2 firsts editions of  Data Science Summer School (DS3), the Ecole polytechnique welcomed from June 24th to June 28th, the third event DS3 on its campus in Palaiseau. This event, made […]

  • Ecological transition : the central theme of the 225th anniversary of the Ecole polytechnique

    Ecole polytechnique decided to celebrate this 225th anniversary by committing to a strategy of sustainable development For its 225th anniversary, Ecole polytechnique organised on June 7, 2019, the international scientific symposium : «RefleXions: researching, educating and acting for sustainable development» […]

  • nam.R was at GeoData Days 2019

    Our Data Strategist team was at GeoData Days 2019 on the behalf of nam.R nam.R was present at the GeoData Days 2019 from july 2nd to july 3rd! Our Data Strategist, Nicolas Berthelot, Alexis Camberlyn and Charles Hutin-Persillon were able […]

  • What you should know about GeodataDays 2018

    The first professional meeting of the Geodata field took place at le Havre on July 3 and 4, 2018: Geodatadays 2018. After they launched each on their side « Rencontres Dynamiques Régionales en information géographique » and « DécryptaGéo» , Afigéo and DécryptaGé […]

  • AI and art: How art and museums are reinventing themselves with data

    In the last few years, AI art has proliferated. Some AI algorithms can recreate an image in a painter’s style, some can improvise a track alongside a human musician on stage and some can even write a story presenting the […]

  • Artificial intelligence transforms travel experience

    In terms of data, tourism industry is complex to analyze. It has just 25% of structured data that comes from varied sources like business websites of tour operators, e-commerce by travel agencies, hosting companies, restaurants, transporters and CRM & other […]

  • Data Science Summer School 2017 – first Edition

    Even before the begining, the international program organised by Emmanuel Bacry (co-founder of nam.R, director of research at the CNRS at Paris Dauphine University and lecturer at Ecole Polytechnique), was a success: 700 applications of which 400 were accepted, 30 […]

  • Data Science Summer School 2018

    An idea of Emmanuel Bacry, co-founder of nam.R, director of research at the CNRS, Paris Dauphine University and lecturer at the Ecole Polytechnique, the Data Science Summer School have gathered this year a new colorful panel of experts. Among them, Cédric Villani, […]

  • Biodiversity and open data: New methods of environmental protection

    The role of big data in biodiversity related issues is rising. A living being leaves such enormous amount of data that it’s difficult to conceive for human mind. This does not frighten the researchers in this field. Following the same […]