1. Introduction
control, anti collision assist, stop and go, overtaking and lane departure warnings. Computer vision is a fundamental part of ADAS and it is used in this project to find and track vehicles in front of the car by processing video streams from a monocular camera. These algorithms are in most cases implemented to run in high-end CPUs of laptops or compact industrial computers to cope with the required computational load. Compared to other sensing modalities such as radars and lasers/lidars, cam- eras are attractive because of the availability of high-quality and low-cost devices and the large amount of information available into a single image. Radars are widely used in anti-collision systems because they are immune to illumination changes and bad weather conditions, they can see far beyond than a camera does but they provide a coarse description of the environment. In contrast, lasers and lidars provide a very detailed map of the environment but they can not sense through dense fog or smog and they are expensive. In order to cope with the shortcomings of individual sensing modalities, intelligent vehicles fuse the information from different sensors. Therefore, the focus is on the hardware and software responsible for receiving and processing sensor data. Modern architectures incorporate notification systems and can react to urgent scenarios to prevent an accident. Nvidia offers intensive data processing solutions to automakers through a visual computing module (VCM), which includes the Tegra family processor designed specifically for the automotive industry [16]. Nvidia also offers a programming language developed for graphical processing units (GPU) providing parallel processing power required to interpret video sequences. Some of these processors are now found in embedded systems, which are critical to guarantee real-time control while being able to withstand vibrations and continuous use. Embedded systems also offer a small size and cost factor when compared to computers. At the Universidad de San Buenaventura Cali, the project Computer-Vision based Driver-Assistance System (ComViDAS) aims to integrate visual perception and self-motion estimation systems to deliver a driver assistance adapted to Colombian roads. The ComViDAS project has proposed the design and development of the following