Luca Mazzei

VIAC: An out of ordinary experiment

This paper presents the preliminary results of VIAC, the VisLab Intercontinental Autonomous Challenge, a test of autonomous driving along an unknown route from Italy to China. It took 3 months to run the entire test; all data have been logged, including all data generated by the sensors, vehicle data, and GPS info. This huge amount of information has been packed during the trip, compressed, and transferred back to Parma for further processing. This data is now ready for a deep analysis of the various systems performance, with the aim of virtually running the whole trip multiple times with improved versions of the software. This paper discusses some preliminary figures obtained by the analysis of the data collected during the test. More information will be generated by a deeper analysis, which will take additional time, being the data about 40 terabyte in size.

The VisLab Intercontinental Autonomous Challenge: An Extensive Test for a Platoon of Intelligent Vehicles

This paper presents the VisLab Intercontinental Autonomous Challenge (VIAC), an autonomous vehicles test carried out from Parma to Shanghai between July and October 2010 by the VisLab team. The vehicle equipment is explained introducing the sensing systems which were tested during the journey. Trip details and the first statistics are presented as well.

Day and night pedestrian detection using cascade AdaBoost system

This paper presents the results of an all-day-long pedestrian classification system based on an AdaBoost cascade meta-algorithm. The underlying idea is to use a Haar-features-based AdaBoost together with an ad-hoc-features-based AdaBoost system in order to reach a better pedestrian classification. A specific night-time pedestrian classification is developed in order to obtain a system that can be used also in poorly illuminated environments. These classifiers are joined together using a cascade AdaBoost system that uses the output of the previous classifiers to obtain a final classification for the area. In the paper the night time and the ad-hoc features systems are presented together with the cascade classification and quantitative results.

Automated extrinsic laser and camera inter-calibration using triangular targets

This paper presents a method for solving the extrinsic calibration between camera and multi-layer laser scanner for outdoor multi-sensorized vehicles. The proposed method is designed for intelligent vehicles within the autonomous navigation task where usually distances between sensor and targets become relevant for safety reasons, therefore high accuracy across different measures must be kept. The calibration procedure takes advantage of triangular shapes still present in scenarios, it recovers three virtual points as target pose in the laser and camera reference frames and then compute extrinsic information of each camera sensor with respect to a laser scanner by minimizing a geometric distance in the image space. To test algorithm correctness, and accuracy a set of simulations are used reporting absolute error results and solution convergence, then tests on robustness and reliability (i.e., outliers management) are based on a wide set of datasets acquired by VIAC prototypes.

A lasers and cameras calibration procedure for VIAC multi-sensorized vehicles

This paper proposes an efficient calibration procedure designed for the vehicle prototypes involved in the VisLab Intercontinental Autonomous Challenge [1]. The perception system is based on laser rangefinders and cameras, for their complementary purpose. An high precision pose estimation must be carried out by a very precise calibration procedure, in order to convert information between different coordinate systems. A modular approach has been extensively tested during VIAC which has offered a unique chance to face pros and cons of different calibration procedures; data collected during the expedition has also become a reference benchmark for further improvements.

Fast vision-based road tunnel detection

When a vehicle equipped with an artificial vision system enters or exits a tunnel, the camera may temporarly suffer from reduced visibility, or even get completely blind due to quick changes in enviromental illumination. This paper presents a vision-based system that detects approaching tunnels entrances or exits. The proposed system allows other ADAS (Advanced Driver Assistance Systems) to act on camera parameters to effectively avoid the tunnel blindness effect. Information regarding approaching tunnel entrance can be helpful for other sensors as well and for sensor fusion systems. In terms of path planning, this system can also inform GNSS-based systems (Global Navigation Satellite System), which usually do not receive any signal in tunnels, and trigger dead reckoning techniques. The proposed system is noticeably fast and therefore well fit to be used as a background process to support other ADAS applications.

Comparison of Three Approaches for Scenario Classification for the Automotive Field

To extend the functionalities of Advanced Driver Assistance Systems (ADAS) and have a more accurate control on the parameters of sensors mounted on an intelligent vehicle, a tool that can classify the scenarios which the vehicle moves in, is needed.