Research Results

2014 Vehicle and pedestrian detection

MultiClass Object Detection at 30fps on Single CPU. Training set INRIA/KITTI. Test on VisLab ETH-ZURICH.

2014 Structure From Motion and Visual Odometry

I will update shortly

2013 Autonomous vehicle control

I will update shortly

2007 Traffic Sign Recognition

I will update shortly

2007 Obstacle Detection using arbitrary aligned stereo cameras

Using low cost camera, a pair images are both distortion and perspective corrected and consequentely matched. This technique are used to provide warning to driver in blind spot application for Heavy Good Vehicles (HGV) and automatic driving for autonomous vehicles. Presented in Versailles, Paris in 2007 for final demostration of Prevent-Apalaci Project.

2007 Obstacle detection and classification with fusion of single camera and radar

Using a Motion Stereo technique vertical and moving objects are detected using a single, monocular, b/w camera. Instantaneous camera calibration for motion stereo is provided using a lane detection technique. Vehicles are detected using fusion between a 24Ghz radar and the same camera, and pedestrian detected using a naive simmetry detector fused with motion stereo information. Presented in Versailles, Paris in 2007 for final demostration of Prevent-Apalaci Project.

2007 Obstacle detection using Far Infrared (FIR) camera in nautical environment

This demo shows obstacles detection technique based on Far Infrared images, able to detect dangerous obstacle in sea (boat, float, and so on), using only a single camera, estimating roughly the distance of the object.

2004 Path Detection for unstructured environment

Unstructured road detection developed for Team Oshkosh, Grand Challenge 2004. Based on color local analysis of image provided by single camera.