Smart drone imagery for the management of tropical forest ecosystems - I-DROP

The I-drop project aims to provide innovative solutions for the start-up Sunbirds in the context of the development of its industrial activities. The goal is to design procedures to identify commercial tree species in the Congo Basin using drones.
© A. Rival, Cirad
© A. Rival, Cirad

© A. Rival, CIRAD


Management plans in the Central African forests require inventories of trees prior to logging. These forest inventories are conducted by teams of prospectors and are long and time-consuming. This is why practical solutions are being sought. In this context, the company Interholco contacted the start-up Sunbirds, which specialises in the deployment of long-range drones. Based on this industrial partnership, a scientific consortium was then formed, involving CEA, CIRAD, Marien Ngouabi University and Nature+. For Sunbirds, the challenge is to expand its range of products based on drones equipped with sensors at the visible and infrared wavelengths.


In order to improve the recognition of commercial tree species in the tropical forests of Central Africa, scientists are working to develop smart solutions with machine learning. Drone technology is enabling them to produce large field data sets, which are essential to detect rare species.
The company Sunbirds has developed a long-range drone that can fly for eight hours, and thus cover vast territories. This technology is therefore perfectly suited to forest concessions, which are very large. In order to align technology products with operational demand, the goal was to develop a tool to analyse data obtained by drones for the identification of commercial trees to be inventoried.
The project focused on obtaining a smart model to identify five of the most sought-after tree species. The shape, spectral signature and structure of the crown of each species were catalogued tree by tree in the training area. This database enabled the model developed by machine learning to learn to recognise all trees from each species over very large areas. The model is “open”, in other words new species can be added to it if the keys to their identification are provided.

Expected impacts

At the end of the project, Sunbirds will be able to expand through new deals. In addition, logging companies will be able to conduct and update inventories in a more flexible, rapid and cost-effective manner.
At the scientific level, this innovative work will open up research on the operational use of drones in tropical forests, and will develop machine learning approaches.


Contract partners   

  • Sunbirds company
  •  Interholco timber company (Congo)
  • French Alternative Energies and Atomic Energy Commission (CEA) (coordinator)
  • Marien Ngouabi University, Brazzaville (Congo)
  • Nature+ NGO (Belgium)