Yann Goyat

Vehicle trajectories evaluation by static video sensors

Metrology of vehicle trajectories has several applications in the field of road safety, particularly in dangerous curves. Actually, it is of great interest to observe trajectories of vehicles with the aim of designing a real time driver warning device in dangerous areas. This paper addresses the first step of a work with a video system placed along the road with the objective of vehicles position and speed estimation. This system has been totally developed for this project and can record simultaneously three cameras with 640 times 480 pixels up to 30 frames per second (fps) and rangefinder informations. The best contribution of this paper is an original probabilistic background subtraction algorithm, first step of a global method (calibration, tracking, ...) implemented to be able to measure vehicle trajectories. Kinematic GPS (in post-processing) has been extensively used to get ground truth

A benchmark for Background Subtraction Algorithms in monocular vision: A comparative study

Background subtraction of video sequences is mainly regarded as a solved problem. However, no complete benchmark about Background Subtraction Algorithms (BSA) has been established, with ground truth and associated quality measures. One of the reasons is that such comparative study needs annotated datasets. In this article, we propose a BSA evaluation dataset built from realistic synthetic image and we compare six BSA, according to several quality measures.

Tracking of vehicle trajectory by combining a camera and a laser rangefinder

This article presents a probabilistic method for vehicle tracking using a sensor composed of both a camera and a laser rangefinder. Two main contributions will be set forth in this paper. The first involves the definition of an original likelihood function based on the projection of simplified 3D vehicle models. We will also propose an efficient approach to compute this function using a line-based integral image. The second contribution focuses on a sampling algorithm designed to handle several sources. The resulting modified particle filter is capable of naturally merging several observation functions in a straightforward manner. Many trajectories of a vehicle equipped with a kinematic GPS1 have been measured on actual field sites, with a video system specially developed for the project. This field input has made it possible to experimentally validate the result obtained from the algorithm. The ultimate goal of this research is to derive a better understanding of driver behavior in order to assist road managers in their effort to ensure network safety.

Vehicle trajectory estimation using spatio-temporal MCMC

This paper presents an algorithm for modeling and tracking vehicles in video sequences within one integrated framework. Most of the solutions are based on sequential methods that make inference according to current information. In contrast, we propose a deferred logical inference method that makes a decision according to a sequence of observations, thus processing a spatio-temporal search on the whole trajectory. One of the drawbacks of deferred logical inference methods is that the solution space of hypotheses grows exponentially related to the depth of observation. Our approach takes into account both the kinematic model of the vehicle and a driver behavior model in order to reduce the space of the solutions. The resulting proposed state model explains the trajectory with only 11 parameters. The solution space is then sampled with a Markov Chain Monte Carlo (MCMC) that uses a model-driven proposal distribution in order to control random walk behavior. We demonstrate our method on real video sequences from which we have ground truth provided by a RTK GPS (Real-Time Kinematic GPS). Experimental results show that the proposed algorithm outperforms a sequential inference solution (particle filter).

Transductive Transfer Learning to Specialize a Generic Classifier Towards a Specific Scene

In this paper, we tackle the problem of domain adaptation to perform object-classification and detection tasks in video surveillance starting by a generic trained detector. Precisely, we put forward a new transductive transfer learning framework based on a sequential Monte Carlo filter to specialize a generic classifier towards a specific scene. The proposed algorithm approximates iteratively the target distribution as a set of samples (selected from both source and target domains) which feed the learning step of a specialized classifier. The output classifier is applied to pedestrian detection into a traffic scene. We have demonstrated by many experiments, on the CUHK Square Dataset and the MIT Traffic Dataset, that the performance of the specialized classifier outperforms the generic classifier and that the suggested algorithm presents encouraging results.

M2SIR: A multi modal sequential importance resampling algorithm for particle filters

We present a multi modal sequential importance resampling particle filter algorithm for object tracking. We consider a hidden state sequence linked to several observation sequences given by different sensors. In a particle filter based framework, each sensor provides a likelihood (weight) associated to each particle and simple rules are applied to merge the different weights such as addition or product. We propose an original algorithm based on likelihood ratios to merge the observations within the sampling step. The algorithm is compared with classic fusion operations on toy examples. Moreover, we show that the method gives satisfactory results on a real vehicle tracking application.