Les commentaires sont clos.

M2 internship: 3D modelling of a fire plume from drone vision

21 Octobre 2022

Catégorie : Stagiaire

Context: The availability of information on a smoke plume is essential to understand the characteristics of a fire, and also to predict the pollution caused. In particular, the geometric shape of the plume and its evolution over time is important to know.



The aim of the internship is to define and integrate the set of algorithms that would allow the reconstruction of this 3D geometrical shape on the basis of image sequences acquired by a drone. The problem is made difficult by the visual aspect of a plume, but especially by the fact that it is a phenomenon that evolves rapidly. We will study the adaptation of classical algorithms for the detection and tracking of visual points of interest and the joint reconstruction of the position and speed of movement of these points. This problem being largely underdetermined, we will investigate how a priori assumptions can be used to solve it. Finally, even if it is not planned to integrate the algorithms on board a drone during the internship, we will privilege approaches that can be executed in real-time without large computational resources.

The work will be carried out incrementally and will start with the exploitation of image sequences acquired on the ground by different cameras to facilitate the 3D reconstruction problem. We will possibly also consider the exploitation of infrared images.

Desired profile:

The candidate must have a degree in image processing and geometry of vision (calibration techniques, stereovision, 3D reconstruction), and know-how in applied mathematics in general. Mastery of Python and C/C++ is required, preferably under Linux, knowledge of OpenCV is a plus.


Applications (CV and cover letter) should be sent to and The internship will last 4 to 6 months and will be carried out at LAAS-CNRS in the Robotics and Interaction team (RIS). The intern will receive a monthly allowance of about 550 €. Applications whose CV does not match these skills or whose cover letter is not adapted to the proposed topic will not be considered.


[1] Crispel, P., & Roberts, G. (2018). All-sky photogrammetry techniques to georeference a cloud field. Atmospheric measurement techniques, 11(1), 593-609.
[2] Martinez-de Dios, J. R., Arrue, B. C., Ollero, A., Merino, L., & Gomez-Rodrıguez, F. (2008). Computer vision techniques for forest fire perception. Image and Vision Computing, 26, 550-562.