PhD Student in Applied Mathematics

Public institution: Cerema, direction Normandie-Centre, Le Grand-Quevilly

Team: Research Team ENDSUM (Non-Destructive Evaluation of Structures and Materials), INSA Rouen Normandie, LMI, EA 3226 – FR CNRS 3335

Contact and supervision: Carole Le Guyader (INSA Rouen Normandie, Pr.), Raphaël Antoine, PhD and Christophe Heinkelé, PhD (Cerema, Research scientist)

PhD diploma Country: France

University Normandie Université (INSA de Rouen)

Postgraduate school: École Doctorale MIIS

Main objectives

The object of this Ph.D. is the semi-automatic detection of cracks from the combination and/or the fusion of images in the visible and the thermal infrared wavelengths. The purpose is to make a diagnosis on geological materials using digital tools (detection, classification, and segmentation) and then to characterize these cracks (length, estimation of potentially unstable volume, and so on…). A soil movement is a displacement of material induced by stresses natural (heavy rainfall, etc…) or anthropogenic (earthworks, vibration, quarrying, etc.).

The Normandy region is particularly concerned by these events (cavities and cliffs collapses, subsidences). Their manifestation may produce significant damages to structures, buildings and infrastructures. It is thus necessary 1) to monitor such phenomenons to understand how they work and 2) to predict the uncertainty associated to the event, defined by of the volume of unstable soil.

Today, cracks are mapped by geologists on the field. This requires considerable time and may be dangerous. Semi-automatic mapping methods can be used on photos, but remain ineficient when the reflectance of the

crack is similar to its environment. For example, a dark crack will be difficult to detect on a dark background: a contrast is needed. The infrared thermal allows to measure the surface temperatures of an object. For a given material, these temperatures first depend of the orientation of the surfaces to the solar radiation. This is why a crack will not have the same temperature than the surrounding material, due to its geometry (depending on the time of the day).

This method allows a better understanding of these objects, thanks to their distinct thermal behaviour.

We want to use these data specifically to improve detection algorithms. The Ph.D will aim at modifing them with a new reference dataset, richer and of better quality, in order to improve crack extraction. Experimental campaigns will therefore be carried out for that purpose. It will required field or UAV data acquisitions, available at Cerema (Rouen).

This thesis in composed of two parts. The first one relies on the development of a robust method which determines if an image contains a crack or not. The methodology envisaged for the moment is the machine learning (branch of artificial intelligence). The second part  is the precise characterization of a crack detected by 3-dimensional reconstruction (photogrammetry). The scalar field infrared can be directly applied to the 3D model, which means a wide opening to complete diagnosis of geological structures.

Required profile: Applied mathematics, image processing, computer vision, computer science

Deadline for applications: 2019 April 12th

How to apply:

Send a cover letter and a CV to: (+33 2 35 68 90 53) and  (+33 2 32 95 99 14)


About the profile:

Required documents for application

• CV

• Cover letter

• lD or passport copy

• Master scores

• Copy of last degree

• Reference letter