Closing date: 10/01/2022
The objective is to carry out a geometric and photometric modeling of a hyperspectral camera and to implement it in an image synthesis software like blender, rhyno3D or opencascade. The work will start with a bibliographic study and will be followed by the development of the application. The goal of this PhD is to help produce a hyperspectral image database for industrial purposes.
The bibliography work is divided into two themes, hyper-spectral imaging and image synthesis:
• To define the image of an object, it is necessary to know several characteristics such as its reflectance, its shape, its brdf, etc. The bibliographic study will be used to extract the essential characteristics for the generation of hyperspectral images.
• The bibliographic research will also lead to the choice of the image synthesis method (ray tracing, z-buffer, etc.) and the rendering engine best suited to the problem.
Thereafter, the first part of the work will consist in implementing, in the selected software tool, the camera and object modeling (acquisition sensitivity curves, object’s spectral properties, etc.). This will allow us to obtain a set of synthetic images that will be compared to real acquisitions.
In a second part, different noises (from the object, the camera, etc.) can be added to simulate images more realistic. This part will also model defect on the simulated object to create at least 2 classes of object: valid or not.
- Romain Hoarau et al. “Interactive Hyper Spectral Image Rendering on GPU”, in VISIGRAPP, 2018.
- Neil Scanlan et al. “Performance analysis of improved methodology for incorporation of spatial/spectral variability in synthetic hyperspectral imagery.", in Imaging Spectrometry IX, 2004.
- Good programming skills in python or C/C++,
- Good mathematical basics,
- Knowledge of geometric camera modelling,
- Knowledge in colour/multispectral imaging,
- Dynamism and autonomy to integrate a multidisciplinary research team.