**6. Conclusion**

In this chapter we have proposed a processing chain addressing safety at level crossings composed of two steps : a background subtraction based on Spatio-temporal Idependentent Component Analysis and a robust 3D localization by global stereo matching algorithm. It is to be noted that the 3D localization is only applied on stationary and moving obstacles. The foreground extraction method based on stICA has already been evaluated in terms of Recall (95%) and Precision (98%), on a set of 1000 images with manually elaborated ground truth. Real-world datasets have been shot at three different level crossings and a parking at the EPFL institute including a hundred scenarios per level crossing under different illumination and weather conditions. This first step is compared with two well-known robust algorithms, entitled GMM and Codebook, from which it proves it effectiveness in term of precision of foreground extraction and processing time. The stereo matching algorithm is first applied on a standard dataset known as the Middlebury Setreo Vision which represents an unique framework for comparison with the state-of-the-art. The latter proves it effectiveness compared to stereo matching algorithms found in the literature.

The experimentations showed that the method is applicable to real-world scenes in level crossing applications. The main output of the proposed system is an accurate 3D localization of any object in, and around a level crossing. According to the experimentations, the localization of some objects may fail. However, the localization of one among sixty objects fails, this is due to the smaller number of pixels having confidence measure larger than a fixed threshold or the occlusion problem. The starting point of the belief propagation process highly depends on the number and repartition of pixels, having hight confidence measure, inside an object. This drawback can be handled by introducing the temporal dependency in the belief propagation process.

For safety purposes, the proposed system will be coupled with already existing devices at level crossings. For instance, the status of the traffic light and the barriers will be taken as input in our vision-based system. The level of such an alarm depends on the configuration of the different parameters. For instance, the presence of an obstacle in the crossing zone when the barriers are lowering is a dangerous situation and the triggered alarm must be of high importance. A Preliminary Risk Analysis (PRA) seems to be an interesting way to categorize the level of alarms. In the frame of the French project entitled PANSafer, these different parameters will be studied. In particular, telecommunication systems will be used to inform road users on the status of the level crossing. Such informations could also be shared with train driver and control room. The communication tool and the nature of information to be transmitted are in study.
