This project will develop classification models in a multi-class and multicriteria context. These models will be based in part on multicriteria decision-making support techniques and statistical learning techniques. The first stage of the project will develop and validate performance measures of the multiclass classifiers. The second stage will develop classification models based on these performance measures of multi-class classifiers as objective functions of the mathematical model. Solving this type of problem will require developing heuristics.
This project in collaboration with Lockheed Martin Canada and University Laval is on aerial surveillance. The goal is to develop and implement algorithms for the automatic object detection using electro-optics data to be used in conjunction with standard operation with the aim of enhancing the speed and accuracy of the regular detection of objects from aerial images. The challenge is to recognize 3D objects in real-world 2D imagery.
This project aims to design a new intelligent recognition system by combining panoramic imaging and retroflexion techniques. Panoramic imaging allows the user to project hemispheric 360º vision onto a 2D detector. The project then intends to couple the panoramic image with the retroflexion image. Retroflexion imaging consists in emitting a laser pulse and observing with a camera the return of the pulse after its retroflexion on components. The retroflexion principle helps the user identify different components using their optical signature.