Internship proposal: 2012.
Target level: bachelor’s or master's research project.
Effort: not less than 20 hours per week
Remuneration: start from 1500 RMB per month, might be raised up to 2500 RMB p
er month according to his/her effort and output.
Duration: 6 month or more, start from the middle of February.
Contact: Franck DAVOINE (Franck.Davoine@gmail.com), Researcher of the French
National Center for Scientific Research (CNRS), Peking University.
(Franck DAVOINE: http://liama.ia.ac.cn/wiki/user:fdavoine:home)

The internship will be done in Beijing, within the project-team " 3D Mul
timodal Perception and Reasoning " (MPR) of the Sino-French laboratory L
IAMA. The candidate will work in the Key Lab. of Machine Perception (MOE) of
Peking University.
(MPR: http://www.liama.ac.cn/mpr)

Title: Detection and segmentation of planar objects in driving urban scenes.

Introduction: The extraction of planar surfaces like the road or facades in o
ut-door driving environments is an important topic that can contribute to und
erstand the scene, usually composed of static objects (trees, traffic signs,
cars, pedestrians, etc.) and dynamic event (moving objects, interactions betw
een them). Recent studies for road extraction have been proposed based on mon
o- or stereo-cameras and multi-layer laser sensors for road plan extraction.

Schedule arrangement: We will first propose the candidate to study some state
-of-the-art methods for road plane extraction. He will then focus on one of t
hem using stereo disparity maps. In a third time, he will study and develop a
method to segment the road area considering a learning approach and a set of
features such as intensity, color, orientation of gradients or higher level
ones extracted at different scales. The selection of good features is one of
the important goals of the project. Segmentation will have to be robust to ap
pearance variations caused by reflections on the road, illumination variation
s, presence of shadows, physical degradations, marks or paintings, etc. The g
lobal system should be computationally efficient, for a possible integration
in test-bed vehicles of two laboratories (in Beijing and in Compiegne in Fran
ce).

Requirements: good English language abilities both written and spoken, knowle
dge in the fields of machine learning, pattern recognition and/or computer vi
sion. Solid programming skills; the project involves programming in Matlab an
d C/C++

Location: Key Laboratory of Machine Perception (MOE), Peking University.

Preparation: Franck is now in France, and he'll be back on January 5th. You c
an send him your CV and motivations by email, before a possible interview aft
er January 5th.

Keywords: Machine learning, classification/clustering, images, stereo-vision,
motion, driving scenes, intelligent vehicles.
--
FROM 162.105.71.*